Browsing by Subject "Medical imaging"

More than 400,000 anterior cruciate ligament (ACL) injuries occur annually in the United States, 70% of which are non-contact. A severe consequence of ACL injury is the increased risk of early-onset of osteoarthritis (OA). Importantly, the increased risk of OA persists even if the ACL is surgically reconstructed. Thus, due to the long term physical consequences and high financial burden of treatment, injury prevention and improved reconstruction techniques are critical. However, the causes of non-contact ACL injuries remain unclear, which has hindered efforts to develop effective training programs targeted at preventing these injuries. Improved understanding of the knee motions that increase the risk of ACL injury can inform more effective injury prevention strategies. Furthermore, there is presently limited in vivo data to describe the function of ACL under dynamic loading conditions. Understanding how the ACL functions to stabilize the knee joint under physiologic loading conditions can inform design criteria for grafts used in ACL reconstruction. Grafts that more accurately mimic the native function of the ACL may help prevent these severe long term degenerative changes in the knee joint after injury.

To this end, measurements of in vivo ACL function during knee motion are critical to understanding how non-contact ACL injuries occur and the function of the ACL in stabilizing the joint during activities of daily living. Specifically, identifying the knee motions that increase ACL length and strain can elucidate the mechanisms of non-contact ACL injury, as a taut ligament is more likely to fail. Furthermore, measuring ACL elongation patterns during dynamic activity can inform the design criteria for grafts used in reconstructive surgery. To obtain measurements, 3D imaging techniques that can be used to measure dynamic in vivo ACL elongation and strain at high temporal and spatial resolution are needed.

Thus, in this dissertation a method of measuring knee motion and ACL function during dynamic activity in vivo using high-speed biplanar radiography in combination with magnetic resonance (MR) imaging was developed. In this technique, 3D surface models of the knee joint are created from MR images and registered to high-speed biplanar radiographs of knee motion. The use of MR imaging to model the joint allows for visualization of bone and soft tissue anatomy, in particular the attachment site footprints of the ligaments. By registering the bone models to biplanar radiographs using software developed in this dissertation, the relative positions of the bones and associated ligament attachment site footprints at the time of radiographic imaging can be reproduced. Thus, measurements of knee kinematics and ligament function during dynamic activity can be obtained at high spatial and temporal resolution.

We have applied the techniques developed in this dissertation to obtain novel dynamic in vivo measurements of the mechanical function of the knee joint. Specifically, the physiologic elongation and strain behaviors of the ACL during gait and single-legged jumping were measured. Additionally, the dynamic function of the patellar tendon during single legged jumping was measured. The findings of this dissertation have helped to elucidate the knee kinematics that increase ACL injury vulnerability by identifying the dynamic motions that result in elongation and strain in the ACL. Furthermore, the findings of this dissertation have provided critical data to inform design criteria for grafts used in reconstructive surgery such that reconstructive techniques better mimic the physiologic function of the ACL.

The methodologies described in this dissertation can be applied to study the mechanical behavior of other joints such as the spine, and other soft tissues, such as articular cartilage, under various loading conditions. Therefore, these methods may have a significant impact on the field of biomechanics as a whole, and may have applicability to a number of musculoskeletal applications.

Prostate cancer is the most common cancer and second-leading cause of cancer death among men in the United States. Early and accurate diagnosis of prostate cancer remains challenging; following an abnormal rectal exam or elevated levels of prostate-specific antigen in serum, clinical guidelines recommend transrectal ultrasound-guided biopsy. However, lesions are often indistinguishable from noncancerous prostate tissue in conventional B-mode ultrasound images, which have a diagnostic sensitivity of about 30%, so the biopsy is not typically targeted to suspicious regions. Instead, the biopsy systematically samples 12 pre-specified regions of the gland. Systematic sampling often fails to detect cancer during the first biopsy, and while multiparametric MRI (mpMRI) techniques have been developed to guide a targeted biopsy, fused with live ultrasound, this approach remains susceptible to registration errors, and is expensive and less accessible.

The goal of this work is to leverage ultrasound elasticity imaging methods, including acoustic radiation force impulse (ARFI) imaging and shear wave elasticity imaging (SWEI), to develop and optimize a robust 3-D elasticity imaging system for ultrasound-guided prostate biopsies and to quantify its performance in prostate cancer detection. Towards that goal, in this dissertation advanced techniques for generating ARFI and SWEI images are developed and evaluated, and a deep learning framework is explored for multiparametric ultrasound (mpUS) imaging, which combines data from different ultrasound-based modalities.

In Chapter 3, an algorithm is implemented that permits the simultaneous imaging of prostate cancer and zonal anatomy using both ARFI and SWEI. This combined sequence involves using closely spaced push beams across the lateral field of view, which enables the collection of higher signal-to-noise (SNR) shear wave data to reconstruct the SWEI volume than is typically acquired. Data from different push locations are combined using an estimated shear wave propagation time between push excitations to align arrival times, resulting in SWEI imaging of prostate cancer with high contrast-to-noise ratio (CNR), enhanced spatial resolution, and reduced reflection artifacts.

In Chapter 4, a fully convolutional neural network (CNN) is used for ARFI displacement estimation in the prostate. A novel method for generating ultrasound training data is described, in which synthetic 3-D displacement volumes with a combination of randomly seeded ellipsoids are used to displace scatterers, from which simulated ultrasonic imaging is performed. The trained network enables the visualization of in vivo prostate cancer and prostate anatomy, providing comparable performance with respect to both accuracy and speed compared to standard time delay estimation approaches.

Chapter 5 explores the application of deep learning for mpUS prostate cancer imaging by evaluating the use of a deep neural network (DNN) to generate an mpUS image volume from four ultrasound-based modalities for the detection of prostate cancer: ARFI, SWEI, quantitative ultrasound, and B-mode. The DNN, which was trained to maximize lesion CNR, outperforms the previous method of using a linear support vector machine to combine the input modalities, and generates mpUS image volumes that provide clear visualization of prostate cancer.

Chapter 6 presents the results of the first in vivo clinical trial that assesses the use of ARFI imaging for targeted prostate biopsy guidance in a single patient visit, comparing its performance with mpMRI-targeted biopsy and systematic sampling. The process of data acquisition, processing, and biopsy targeting is described. The study demonstrates the feasibility of using 3-D ARFI for guiding a targeted biopsy of the prostate, where it is most sensitive to higher-grade cancers. The findings also indicate the potential for using 2-D ARFI imaging to confirm target location during live B-mode imaging, which could improve existing ultrasonic fusion biopsy workflows.

Chapter 7 summarizes the research findings and considers potential directions for future research. By developing advanced ARFI and SWEI imaging techniques for imaging the prostate gland, and combining information from different ultrasound modalities, prostate cancer and zonal anatomy can be imaged with high contrast and resolution. The findings from this work suggest that ultrasound elasticity imaging holds great promise for facilitating image-guided targeted biopsies of clinically significant prostate cancer.

Purpose: To compare radiomics and deep-learning (DL) methods for predicting NSCLC surgical treatment failure. Methods: A cohort of 83 patients undergoing lobectomy or wedge resection for early-stage NSCLC from our institution was studied. There were 7 local failures and 16 non-local failures (regional and/or distant). Gross tumor volumes (GTV) were contoured on pre-surgery CT datasets after 1mm3 isotropic resolution resampling. For the radiomics analysis, 92 radiomics features were extracted from the GTV and z-score normalizations were performed. The multivariate association between the extracted features and clinical endpoints were investigated using a random forest model following 70%-30% training-test split. For the DL analysis, both 2D and 3D model designs were executed using two different deep neural networks as transfer learning problems: in 2D-based design, 8x8cm2 axial fields-of-view(FOVs) centered within the GTV were adopted for VGG-16 training; in 3D-based design, 8x8x8 cm3 FOVs centered within the GTV were adopted for U-Net’s encoder path training. In both designs, data augmentation (rotation, translation, flip, noise) was included to overcome potential training convergence problems due to the imbalanced dataset, and the same 70%-30% training-test split was used. The performances of the 3 models (Radiomics, 2D-DL, 3D-DL) were tested to predict outcomes including local failure, non-local failure, and disease-free survival. Sensitivity/specificity/accuracy/ROC results were obtained from their 20 trained versions. Results: The radiomics models showed limited performances in all three outcome prediction tasks. The 2D-DL design showed significant improvement compared to the radiomics results in predicting local failure (ROC AUC = 0.546±0.056). The 3D-DL design achieved the best performance for all three outcomes (local failure ROC AUC = 0.768 ± 0.051, non-local failure ROC AUC = 0.683±0.027, disease-free ROC AUC = 0.694±0.042) with statistically significant improvements from radiomics/2D-DL results. Conclusions: 3D-DL execution outperformed the 2D-DL in predicting clinical outcomes after surgery for early-stage NSCLC. By contrast, classic radiomics approach did not achieve satisfactory results.

Diagnosis, prognosis, and treatment of many ocular and neurodegenerative diseases, including achromatopsia (ACHM), require the visualization of microscopic structures in the eye. The development of adaptive optics ophthalmic imaging systems has made high resolution visualization of ocular microstructures possible. These systems include the confocal and split detector adaptive optics scanning light ophthalmoscope (AOSLO), which can visualize human cone and rod photoreceptors in vivo. However, the avalanche of data generated by such imaging systems is often too large, costly, and time consuming to be evaluated manually, making automation necessary. The few currently available automated cone photoreceptor identification methods are unable to reliably identify rods and cones in low-quality images of diseased eyes, which are common in clinical practice.

This dissertation describes the development of automated methods for the analysis of AOSLO images, specifically focusing on cone and rod photoreceptors which are the most commonly studied biomarker using these systems. A traditional image processing approach, which requires little training data and takes advantage of intuitive image features, is presented for detecting cone photoreceptors in split detector AOSLO images. The focus is then shifted to deep learning using convolutional neural networks (CNNs), which have been shown in other image processing tasks to be more adaptable and produce better results than classical image processing approaches, at the cost of requiring more training data and acting as a “black box”. A CNN based method for detecting cones is presented and validated against state-of-the-art cone detections methods for confocal and split detector images. The CNN based method is then modified to take advantage of multimodal AOSLO information in order to detect cones in images of subjects with ACHM. Finally, a significantly faster CNN based approach is developed for the classification and detection of cones and rods, and is validated on images from both healthy and pathological subjects. Additionally, several image processing and analysis works on optical coherence tomography images that were carried out during the completion of this dissertation are presented.

The completion of this dissertation led to fast and accurate image analysis tools for the quantification of biomarkers in AOSLO images pertinent to an array of retinal diseases, lessening the reliance on subjective and time-consuming manual analysis. For the first time, automatic methods have comparable accuracy to humans for quantifying photoreceptors in diseased eyes. This is an important step in the long-term goal to facilitate early diagnosis, accurate prognosis, and personalized treatment of ocular and neurodegenerative diseases through optimal visualization and quantification of microscopic structures in the eye.

Purpose: A deep learning framework based on a conditional Generative Adversarial Network (cGAN) was developed to synthesize high-contrast Magnetization Prepared Rapid Gradient Echo (MP-RAGE) images from common spin-echo MR imaging sequences. This framework utilizes combinations of inputs from T1-weighted (T1-w), T2-weighted (T2-w), and Proton Density-weighted (PD-w) images. The primary objective was to augment the diversity of clinical data by capitalizing on the inherent advantages of MP-RAGE imaging, such as superior contrast, while mitigating its susceptibility to metal artifacts.Methods and Materials: A cGAN image synthesis model, incorporating a U-Net-based generator and a Patch GAN discriminator, was developed. The training was conducted across four distinct configurations, employing combinations of T1-w, T2-w, and PD-w images as inputs to synthesize MP-RAGE images, with and without Proton Density (PD) information, designated as PD-0 and PD-1, respectively. For training, data from 51 patients, comprising 8,160 slices, were used, following a training-to-validation ratio of 90:10. For prediction, data from 14 patients, comprising 2,240 slices, were utilized. The efficacy of the synthesized MP-RAGE images was evaluated using a suite of quantitative metrics, including Mean Absolute Error (MAE), Normalized Cross-Correlation (NCC), Percentage Mutual Information (PMI), and Structural Similarity Index (SSIM). Additionally, a Freesurfer brain segmentation task was performed on both synthesized and ground truth brain images, with the fidelity of synthesized images being indirectly assessed by the calculated Dice coefficient. Results: It was observed that the cGAN-synthesized MP-RAGE images exhibited comparable contrast to the ground truth in the axial view. A decrease in input channel numbers resulted in diminished contrast between certain anatomical structures in the synthetic MP-RAGE images, albeit within an acceptable range. The MAE approached (0.02±0.01), the PMI for two Three-in-One-out synthesis approached(0.76±0.07), the NCC was about (0.91±0.05), and the SSIM was about (0.9±0.1). The Freesurfer segmentation results showed desirable Dice coefficients (mostly above 0.8) for different kinds of inputs, except the One-in-One-out T1-w synthesis. Conclusion: The cGAN framework developed in this study has proven to be a robust and versatile tool for synthesizing high-contrast MP-RAGE images, even in scenarios with single-channel input images. The Freesurfer segmentation results demonstrated that the synthesized MP-RAGE images are highly similar to the ground truth in segmentation tasks, underscoring the potential clinical and research value of the proposed image synthesis model.

Purpose: Ensuring that tumor motion is within the radiation field for high-dose and high-precision radiosurgery in areas greatly influenced by respiratory motion. Therefore tracking the target or gating the radiation beam by using real-time imaging and surrogate motion monitoring methods are employed. However, these methods cannot be used to depict the effect of respiratory motion on tumor deviation. Therefore, an investigation of parameters for method predicting the tumor motion induced by respiratory motion multiple steps ahead of real time is performed. Currently, algorithms exist to make predictions about future real-time events, however these methods are tedious or unable to predict far enough in advance.

Methods and Materials: The algorithm takes data collected from the Varian RPM$ System, which is a one-dimensional (1D) surrogate signal of amplitude versus time. After the 1D surrogate signal is obtained, the algorithm determines on average what an approximate respiratory cycle is over the entire signal using a rising edge function. The signal is further dividing it into three components: (a) training component is the core portion of the data set which is further divided into subcomponents of length equal to the input component; (b) input component serves as the parameter searched for throughout the training component, (c) analysis component used as a validation against the prediction. The prediction algorithm consists of three major steps: (1) extracting top-ranked subcomponents from training component which best-match the input component; (2) calculating weighting factors from these best-matched subcomponents; (3) collecting the proceeding optimal subcomponent and fusing them with assigned weighting factors to form prediction. The prediction algorithm was examined for several patients, and its performance is assessed based on the correlation and root mean square error (RMSE) between prediction and known output.

Results: Respiratory motion data was simulated for 30 cases and 555 patients and phantoms using the RPM system. Simulations were used to optimize prediction algorithm parameters. The simulation cases were used to determine optimal filters for smoothing and number of top-ranked subcomponents to determine optimal subcomponents for prediction. Summed difference results in a value of 0.4770 for the 15 Point Savitzky-Golay filter.

After determining the proper filter for data preprocessing the number of required top-ranked subcomponents for each method was determine. Equal Weighting has a maximum average correlation, c=0.997 when using 1 Subcomponent, Relative Weighting has a maximum average correlation, c=0.997 when using 2 Subcomponents, Pattern Weighting has a maximum average correlation c=0.915 when using 1 subcomponent, Derivative Equal Weighting has a maximum average correlation c=0.976 when using 2 Subcomponents, and Derivative Relative Weighting has a maximum average correlation of c=0.976 when using 5 Subcomponents.

The correlation coefficient and RMSE of prediction versus analysis component distributions demonstrate an improvement during optimization for simulations. This is true for both the full and half cycle prediction. However, when moving to the clinical data the distribution of prediction data, both correlation coefficient and RMSE, there is not an improvement as the optimization occurs. Therefore, a comparison of the clinical data using the 5 Pt moving filter and arbitrarily chosen number of subcomponents was performed. In the clinical data, average correlation coefficient between prediction and analysis component 0.721+/-0.390, 0.727+/-0.383, 0.535+/-0.454, 0.725+/-0.397, and 0.725+/-0.398 for full respiratory cycle prediction and 0.789+/-0.398, 0.800+/-0.385, 0.426+/-0.562, 0.784+/-0.389, and 0.784+/-0.389 for half respiratory cycle prediction for equal weighting, relative weighting, pattern, derivative equal and derivative relative weighting methods, respectively. Additionally, the clinical data average RMSE between prediction and analysis component 0.196+/-0.174, 0.189+/-0.161, 0.302+/-0.162, 0.200+/-0.169, and 0.202+/-0.181 for full respiratory cycle prediction and 0.155+/-0.171, 0.149+/-0.138, 0.528+/-0.179, 0.174+/-0.150, and 0.173+/-0.149 for half respiratory cycle prediction for equal weighting, relative weighting, pattern, derivative equal and derivative relative weighting methods, respectively. The half cycle prediction displays higher accuracy over the full cycle prediction. Wilcoxon signed-rank test reveals statistically highly significant values (p<0.1%) for 4 out of 5 algorithms favoring the half cycle prediction (Equal, Relative, Derivative Equal, and Derivative Relative Weighting Methods). In this method, the relative weighting method has the most correlations coefficients with values greater than 0.9 and also yields the largest number of highest correlations over other prediction methods.

Conclusions: In conclusion, the number of subcomponents used for prediction may be better determined based on individual breathing pattern. The prediction accuracy using patient data is better using half cycle prediction over full cycle prediction for all algorithms for the majority of methods tested. Finally, relative weighting method performed better than other methods.

AbstractPinhole imaging is a widely used method for high spatial resolution single gamma imaging with a small required field of view (FOV). Many factors affect pinhole imaging: (I) the geometric parameters of the pinhole imaging system, such as the pinhole diameter, focal length and opening angle; (II) the position, range, sampling interval, and sampling time of the pinhole trajectory; and (III) the image reconstruction algorithm. These differences result in different trade-offs between resolution, sensitivity, noise level, imaging FOV, and data-sampling integrity levels. In pinhole imaging, many different pinhole trajectories might be considered. The conventional approach to assessing different trajectories is to reconstruct images from the various trajectories and then assess which image is best. Such an approach is however time consuming, since (I) image reconstruction is time-consuming and (II) image analysis often requires ensembles of images, where the ensemble is time consuming to calculate, consumes considerable computer storage, and requires investigator time to organize and analyze. The object of this project is to develop a method to rapidly select the optimal SPECT pinhole trajectory from among several candidate trajectories. Equivalent Resolution Geometric Efficiency (ERGE) is proposed to represent the spatial resolution and geometric efficiency; a higher ERGE means a better trajectory. To verify this metric, two-dimensional and three-dimensional visualizations of the pinhole trajectory are implemented in software as a way to assess trajectories visually and qualitatively. Several different trajectories are employed, projection data are computer-simulated, including spatial resolution blurring and pseudo-random Poisson noise, and image reconstruction is performed using the OSEM algorithm. The reconstructed images are analyzed to characterize the performance of the different trajectories to assess whether the best trajectory can be determined by the sensitivity and resolution characteristics of the individual pinhole locations that make up the trajectory. Ultimately, the method proved to be effective. In this study, a relatively simple low-cost prospective method for selecting the optimal SPECT pinhole trajectory has been shown to be effective. Only very fast and simple calculations, utilizing Microsoft Excel for example, are required. The method does not require simulating or acquiring projection data and does not require image reconstruction. The ranking of ERGE matches well with the ranking of reconstructed images based on Root Mean Square Error (RMSE). In clinical and scientific research, many different pinhole trajectories might be considered for pinhole 3D SPECT imaging, but it is too time-consuming to assess each trajectory via reconstructed images. By demonstrating the validity of this method for assessing trajectories, it may facilitate the improved use of 3D pinhole SPECT imaging in clinical and scientific research. Keywords: Pinhole Trajectory, SPECT, Forward Projection, OSEM, Equivalent Resolution Geometric Efficiency (ERGE), Root Mean Square Error (RMSE).

Purpose: To predict post-radiotherapy overall survival group of NSCLC patients based on clinical information and radiomics analysis of simulation CT. Materials/Methods: A total of 258 non-adenocarcinoma patients who received radical radiotherapy or chemo-radiation were studied: 45/50/163 patients were identified as short(0-6mos)/mid(6-12mos)/long(12+mos) survival groups, respectively. For each patient, we first extracted 76 radiomics features within the gross tumor volume(GTV) identified in the simulation CT; these features were combined with patient clinical information (age, overall stage, and GTV volume) as a patient-specific feature vector, which was utilized by a 2-step machine learning model for survival group prediction. This model first identifies patients with long survival prediction via a supervised binary classifier; for those with otherwise prediction, a 2nd classifier further generates short/mid survival prediction. Two machine learning classifiers, explainable boosting machine(EBM) and balanced random forest(BRF), were interrogated as a comparison study. During the model training, all patients were divided into training/test sets by an 8:2 ratio, and 100-fold random sampling were applied to the training set with a 7:1 validation ratio. Model performances were evaluated by the sensitivity, accuracy, and ROC results. Results: The model with EBM demonstrated an overall ROC AUC (0.58±0.04) with limited sensitivities in short (0.02±0.04) and mid group (0.11±0.08) predictions due to imbalanced data sample distribution. In contrast, the model with BRF improved short/mid group sensitivities to 0.32±0.11/0.29±0.16, respectively, but the improvement of ROC AUC (0.60±0.04) is limited. Nevertheless, both EBM (0.46±0.04) and BRF (0.57±0.04) approaches achieved limited overall accuracy; a noticeable overlap was found in their feature lists with top 10 feature weight rankings. Conclusion: The proposed two-step machine learning model with BRF classifier possesses a better performance than the one with EBM classifier in the post-radiotherapy survival group prediction of NSCLC. Future works, preferably in the joint use of deep learning, are in demand to further improve the prediction results.

Purpose: To develop a radiomics-embedded vision transformer (RE-ViT) model by incorporating radiomics features into its architecture, seeking to improve the model's efficiency in medical image recognition towards enhanced breast ultrasound image diagnostic accuracy.Materials and Methods: Following the classic ViT design, the input image was first resampled into multiple 16×16 grid image patches. For each patch, 56-dimensional habitat radiomics features, including intensity-based, Gray Level Co-Occurrence Matrix (GLCOM)-based, and Gray Level Run-Length Matrix (GLRLM)-based features, were extracted. These features were designed to encode local-regional intensity and texture information comprehensively. The extracted features underwent a linear projection to a higher-dimensional space, integrating them with ViT’s standard image embedding process. This integration involved an element-wise addition of the radiomics embedding with ViT’s projection-based and positional embeddings. The resultant combined embeddings were then processed through a Transformer encoder and a Multilayer Perceptron (MLP) head block, adhering to the original ViT architecture. The proposed RE-ViT model was studied using a public BUSI breast ultrasound dataset of 399 patients with benign, malignant, and normal tissue classification. The comparison study includes: (1) RE-ViT versus classic ViT training from scratch, (2) pre-trained RE-ViT versus pre-trained ViT (based on ImageNet-21k), (3) RE-ViT versus VGG-16 CNN model. The model performance was evaluated based on accuracy, ROC AUC, sensitivity, and specificity with 10-fold Monte-Carlo cross validation. Result: The RE-ViT model significantly outperformed the classic ViT model, demonstrating superior overall performance with accuracy = 0.718±0.043, ROC AUC = 0.848±0.033, sensitivity = 0.718±0.059, and specificity = 0.859±0.048. In contrast, the classic ViT model achieved accuracy = 0.473±0.050, ROC AUC = 0.644±0.062, sensitivity = 0.473±0.101, and specificity = 0.737±0.065. Pre-trained versions of RE-ViT also showed enhanced performance (accuracy = 0.864±0.031, ROC AUC = 0.950±0.021, sensitivity = 0.864±0.074, specificity = 0.932±0.036) compared to pre-trained ViT (accuracy = 0.675±0.111, ROC AUC = 0.872±0.086, sensitivity = 0.675±0.129, specificity = 0.838±0.096). Additionally, RE-ViT surpassed VGG-16 CNN results (accuracy = 0.553±0.079, ROC AUC = 0.748±0.080, sensitivity = 0.553±0.112, specificity = 0.777±0.089). Conclusion: The proposed radiomics-embedded ViT was successfully developed for ultrasound-based breast tissue classification. Current results underscore the potential of our approach to advance other transformer-based medical image diagnosis tasks.

AbstractPurpose: To develop a deep ensemble learning model with a radiomics spatial encoding execution for improved glioma segmentation accuracy using multi-parametric MRI (mp-MRI). Materials/Methods: This radiomics-incorporated deep ensemble learning model was developed using 369 glioma patients with a 4-modality mp-MRI protocol: T1, contrast-enhanced T1 (T1-Ce), T2, and FLAIR. In each modality volume, a 3D sliding kernel was implemented across the brain to capture image heterogeneity: fifty-six radiomic features were extracted within the kernel, resulting in a 4th order tensor. Each radiomic feature can then be encoded as a 3D image volume, namely a radiomic feature map (RFM). For each patient, all RFMs extracted from all 4 modalities were processed by the Principal Component Analysis (PCA) for dimension reduction, and the first 4 principal components (PCs) were selected. Next, four deep neural networks following the U-net’s architecture were trained for the segmenting of a region-of-interest (ROI): each network utilizes the mp-MRI and 1 of the 4 PCs as a 5-channel input for 2D execution. Last, the 4 softmax probability results given by the U-net ensemble were superimposed and binarized by Otsu’s method as the segmentation result. Three deep ensemble models were trained to segment enhancing tumor (ET), tumor core (TC), and whole tumor (WT), respectively. Segmentation results given by the proposed ensemble were compared to the mp-MRI-only U-net results. Results: All 3 radiomics-incorporated deep learning ensemble models were successfully implemented: Compared to mp-MRI-only U-net results, the dice coefficients of ET (0.777→0.817), TC (0.742→0.757), and WT (0.823→0.854) demonstrated improvements. Accuracy, sensitivity, and specificity results demonstrated the same patterns. Conclusion: The adopted radiomics spatial encoding execution enriches the image heterogeneity information that leads to the successful demonstration of the proposed neural network ensemble design, which offers a new tool for mp-MRI-based medical image segmentation.

The main objective of this dissertation research is to develop measurement and data-analysis tools for improving the quantitative accuracy of radiotracer studies of small plants, e.g., grasses in their early growth stages and tree seedlings. Improved accuracy is needed due to the thin nature of plant organs, e.g., leafs and stem. In addition, the methods developed in this thesis are applied to study the plant-environment interface of barley. Some of the approaches explored have potential to increase the statistical accuracy of counting data using PET imaging techniques. Improving the statistical precision of radionuclide tracking data will add to the analysis options. Another important goal is to measure the absolute photosynthetic rate. The standard approach in plant radiotracer experiments is to perform measurements of the relative distribution of radioactivity in various parts of the plant being studied. A limitation of this approach is that it does not take into account differences in the amount of radioisotope assimilated that are available for transport and allocation to the various sinks, that is, absolute CO2 uptake and photosynthetic rates are important factors in understanding the holistic physiological responses of plants to external conditions. For example, monitoring the movement of carbon-11 (11C) tagged carbohydrates in a plant requires an estimate of the average photosynthetic rate to determine the actual quantity of carbohydrates in each plant region (e.g. leaf, shoot, and root).

Radiotracing provides a method for real-time measurements of substance absorption, allocation and metabolic consumption and production in living organisms. Application of radioactive labelling in plants enables measurements associated with core physiological processes, e.g., photosynthesis, water uptake and nitrogen absorption and utilization. Plant uptake of radiotracers allows for tracking spatial and temporal distribution of substances, which enables studies of the plant-environment interface and the mechanisms involved in the allocation of resources (e.g., sugars, nutrients, and water). As such, these techniques are increasingly becoming an important tool for investigating the processes involved in the physiological responses of plants to changes in their local environmental conditions.

This dissertation has two major components: (1) development of experiment techniques for absolute photosynthetic rate measurements in plants using radio-isotope labeling, and (2) application of radioisotope tracing techniques to study the plant-environment interface in barley. The first component is covered in chapters one through three. The second component is presented in chapter four. An introduction into radio-tracing techniques is provided in chapter one. Chapter two describes radio-isotope production, radio-labelled compound preparation and delivery of labels to plant measurements. Chapter three outlines methods that can be employed to measure the absolute photosynthetic rate (µmol/m2/s) for a closed-loop system with [CO2] monitoring capabilities. Chapter four describes the background and results of our study on changing environmental conditions on a model system, barley seedlings. Chapter 5 will introduce the use of Monte-Carlo modeling for scaling the collected data to adjust the detected coincidence counts for losses due to positron escape from plant tissue. Chapter 6 describes the development of a novel imaging technique using direct positron detection that takes advantage of the high fraction of positrons escaping thin plant tissue.

In this dissertation, we have performed the most extensive measurements of carbohydrate allocation and translocation in a plant species using radio-isotope tracing techniques. A major practical limitation of studies based on radio-isotope labeling is the number of samples that can be measured in a single project. Our study on barley (Hordeum distichum) includes measurements on more than 30 plants. The short-lived radionuclide, 11C, was used to determine the real-time response to metabolite transport in barley. Sugars are photosynthesized and tagged with a positron-emitting radioisotope by flowing carbon dioxide (11CO2) tagged air over an active leaf. Data analysis of measurements taken in this dissertation indicates that the fraction of carbohydrates allocated to below ground sinks decreased, by 31% ± 9% in ambient [CO2] and by 37% ± 14% in elevated [CO2], when the nutrient conditions were rapidly changed from high to low nutrient.

There always exists a demand to accelerate the time-consuming MRI acquisition process. Many methods have been proposed to achieve this goal, including deep learning method which appears to be a robust tool compared to conventional methods. While many works have been done to evaluate the performance of neural networks on standard anatomical MR images, few attentions have been paid to accelerating other less conventional MR image acquisitions. This work aims to evaluate the feasibility of neural networks on accelerating Brain DTI and Gynecological Brachytherapy MRI. Three neural networks including U-net, Cascade-net and PD-net were evaluated. Brain DTI data was acquired from public database RIDER NEURO MRI while cervix gynecological MRI data was acquired from Duke University Hospital clinic data. A 25% Cartesian undersampling strategy was applied to all the training and test data. Diffusion weighted images and quantitative functional maps in Brain DTI, T1-spgr and T2 images in GYN studies were reconstructed. The performance of the neural networks was evaluated by quantitatively calculating the similarity between the reconstructed images and the reference images, using the metric Total Relative Error (TRE). Results showed that with the architectures and parameters set in this work, all three neural networks could accelerate Brain DTI and GYN T2 MR imaging. Generally, PD-net slightly outperformed Cascade-net, and they both outperformed U-net with respect to image reconstruction performance. While this was also true for reconstruction of quantitative functional diffusion weighted maps and GYN T1-spgr images, the overall performance of the three neural networks on these two tasks needed further improvement. To be concluded, PD-net is very promising on accelerating T2-weighted-based MR imaging. Future work can focus on adjusting the parameters and architectures of the neural networks to improve the performance on accelerating GYN T1-spgr MR imaging and adopting more robust undersampling strategy such as radial undersampling strategy to further improve the overall acceleration performance.

This dissertation introduces and analyzes algorithms that aim to adaptively handle complex datasets arising in the real-world applications. It contains two major parts. The first part describes an adaptive model of 1-dimensional signals that lies in the field of adaptive time-frequency analysis. It explains a current state-of-the-art work, named the Synchrosqueezed transform, in this field. Then it illustrates two proposed algorithms that use non-parametric regression to reveal the underlying os- cillatory patterns of the targeted 1-dimensional signal, as well as to estimate the instantaneous information, e.g., instantaneous frequency, phase, or amplitude func-

tions, by a statistical pattern driven model.

The second part proposes a population-based imaging technique for human brain

bundle/connectivity recovery. It applies local streamlines as novelly adopted learn- ing/testing features to segment the brain white matter and thus reconstruct the whole brain information. It also develops a module, named as the streamline diffu- sion filtering, to improve the streamline sampling procedure.

Even though these two parts are not related directly, they both rely on an align- ment step to register the latent variables to some coordinate system and thus to facilitate the final inference. Numerical results are shown to validate all the pro- posed algorithms.

Purpose: The purpose of this study is to develop and validate two deep learning (DL) models for the management of brain metastasis (BM) patients treated with stereotactic radiosurgery (SRS). The first model is a radiomics-integrated deep learning (RIDL) model, which aims to distinguish between radionecrosis and tumor recurrence in patients with post-SRS radiographic progression. The second model, a novel dose-incorporated deep ensemble learning (DEL) model, aims to accurately predict local failure outcomes in brain metastasis patients following SRS.Materials/Methods: A total of 51 patients with post-SRS radiographic progression (37 radionecrosis, 14 recurrence) and 114 BMs (including 26 BMs that developed biopsy-confirmed local failure post-SRS) from 85 patients were included in this study. For the first aim, a radiomics-integrated deep learning (RIDL) model was developed using three steps: 1) 184 radiomics features (RFs) were extracted from the SRS planning target volume (PTV) and 60% isodose volume (V60%); 2) a deep neural network (DNN) mimicking the encoding path of U-net was trained for radionecrosis or recurrence prediction using the 3D MR volume. Prior to the binary prediction output, latent variables in the DNN were extracted as 512 deep features (DFs); and 3) all extracted features were synthesized as a multi-dimensional input for support vector machine (SVM) execution. Key features with higher linear kernel weighting factors were identified by clustering analysis and were utilized by SVM to predict radionecrosis or recurrence. During model training, 50 model versions were acquired with random validation sample assignments following an 8:2 training/test ratio, and sensitivity, specificity, accuracy, and ROC were evaluated and compared with results from a radiomics-only and a DNN-only prediction model. For the second aim, a novel dose-incorporated deep ensemble learning (DEL) model was developed. The DEL design included four VGG-19 deep encoder networks, and each sub-network utilized a different variable type as input for BM outcome prediction. The DEL's outcome was synthesized from the four sub-network results via logistic regression. For each BM, four variables were obtained, including three with different curvatures during spherical projection and one with the original planar images. The proposed DEL model was developed using an 8:2 ratio for training/test assignment, and 10 model versions were acquired with random validation sample assignments. The DEL model performance was compared based on ROC analysis to a single VGG-19 encoder and to DEL models with the same projection designs, which used T1-CE MRI as the only input. Results: The RIDL model demonstrated superior performance compared to radiomics-based and DNN-only prediction models for distinguishing radionecrosis from tumor recurrence in brain metastasis patients with post-SRS radiographic progression. The RIDL model achieved the best prediction accuracy (0.643±0.059) and sensitivity (0.650±0.122) results with 32 identified key features (3 RFs+29 DFs), and it also demonstrated superior ROC results (AUC=0.688±0.035). In addition, for patients with NSCLC primary disease, the RF joint energy extracted from V60% and one DF correlated with ALK/EGFR mutations, respectively. Moreover, the DEL model achieved an excellent ROC AUC=0.84±0.03 with high sensitivity (0.78±0.08), specificity (0.81±0.09), and accuracy (0.80±0.06) results. This outperformed the MRI-only single VGG-19 encoder (sensitivity:0.35±0.01, AUC:0.64±0.08) and the MRI-only DEL (sensitivity:0.60±0.09, AUC:0.68±0.06) models. Conclusions: The RIDL model successfully differentiates brain metastasis radionecrosis from recurrence using a single post-SRS MR scan. Integration of clinical and treatment-related features is warranted to develop a comprehensive clinico-radiomic model. Additionally, the dose-incorporated DEL model design demonstrated robust and promising performance. It could potentially improve other radiotherapy outcome models and warrant further evaluation.

X-ray computed tomography (CT) imaging constitutes one of the most widely used diagnostic tools in radiology today with nearly 85 million CT examinations performed in the U.S in 2011. CT imparts a relatively high amount of radiation dose to the patient compared to other x-ray imaging modalities and as a result of this fact, coupled with its popularity, CT is currently the single largest source of medical radiation exposure to the U.S. population. For this reason, there is a critical need to optimize CT examinations such that the dose is minimized while the quality of the CT images is not degraded. This optimization can be difficult to achieve due to the relationship between dose and image quality. All things being held equal, reducing the dose degrades image quality and can impact the diagnostic value of the CT examination.

A recent push from the medical and scientific community towards using lower doses has spawned new dose reduction technologies such as automatic exposure control (i.e., tube current modulation) and iterative reconstruction algorithms. In theory, these technologies could allow for scanning at reduced doses while maintaining the image quality of the exam at an acceptable level. Therefore, there is a scientific need to establish the dose reduction potential of these new technologies in an objective and rigorous manner. Establishing these dose reduction potentials requires precise and clinically relevant metrics of CT image quality, as well as practical and efficient methodologies to measure such metrics on real CT systems. The currently established methodologies for assessing CT image quality are not appropriate to assess modern CT scanners that have implemented those aforementioned dose reduction technologies.

Thus the purpose of this doctoral project was to develop, assess, and implement new phantoms, image quality metrics, analysis techniques, and modeling tools that are appropriate for image quality assessment of modern clinical CT systems. The project developed image quality assessment methods in the context of three distinct paradigms, (a) uniform phantoms, (b) textured phantoms, and (c) clinical images.

The work in this dissertation used the “task-based” definition of image quality. That is, image quality was broadly defined as the effectiveness by which an image can be used for its intended task. Under this definition, any assessment of image quality requires three components: (1) A well defined imaging task (e.g., detection of subtle lesions), (2) an “observer” to perform the task (e.g., a radiologists or a detection algorithm), and (3) a way to measure the observer’s performance in completing the task at hand (e.g., detection sensitivity/specificity).

First, this task-based image quality paradigm was implemented using a novel multi-sized phantom platform (with uniform background) developed specifically to assess modern CT systems (Mercury Phantom, v3.0, Duke University). A comprehensive evaluation was performed on a state-of-the-art CT system (SOMATOM Definition Force, Siemens Healthcare) in terms of noise, resolution, and detectability as a function of patient size, dose, tube energy (i.e., kVp), automatic exposure control, and reconstruction algorithm (i.e., Filtered Back-Projection– FPB vs Advanced Modeled Iterative Reconstruction– ADMIRE). A mathematical observer model (i.e., computer detection algorithm) was implemented and used as the basis of image quality comparisons. It was found that image quality increased with increasing dose and decreasing phantom size. The CT system exhibited nonlinear noise and resolution properties, especially at very low-doses, large phantom sizes, and for low-contrast objects. Objective image quality metrics generally increased with increasing dose and ADMIRE strength, and with decreasing phantom size. The ADMIRE algorithm could offer comparable image quality at reduced doses or improved image quality at the same dose (increase in detectability index by up to 163% depending on iterative strength). The use of automatic exposure control resulted in more consistent image quality with changing phantom size.

Based on those results, the dose reduction potential of ADMIRE was further assessed specifically for the task of detecting small (<=6 mm) low-contrast (<=20 HU) lesions. A new low-contrast detectability phantom (with uniform background) was designed and fabricated using a multi-material 3D printer. The phantom was imaged at multiple dose levels and images were reconstructed with FBP and ADMIRE. Human perception experiments were performed to measure the detection accuracy from FBP and ADMIRE images. It was found that ADMIRE had equivalent performance to FBP at 56% less dose.

Using the same image data as the previous study, a number of different mathematical observer models were implemented to assess which models would result in image quality metrics that best correlated with human detection performance. The models included naïve simple metrics of image quality such as contrast-to-noise ratio (CNR) and more sophisticated observer models such as the non-prewhitening matched filter observer model family and the channelized Hotelling observer model family. It was found that non-prewhitening matched filter observers and the channelized Hotelling observers both correlated strongly with human performance. Conversely, CNR was found to not correlate strongly with human performance, especially when comparing different reconstruction algorithms.

The uniform background phantoms used in the previous studies provided a good first-order approximation of image quality. However, due to their simplicity and due to the complexity of iterative reconstruction algorithms, it is possible that such phantoms are not fully adequate to assess the clinical impact of iterative algorithms because patient images obviously do not have smooth uniform backgrounds. To test this hypothesis, two textured phantoms (classified as gross texture and fine texture) and a uniform phantom of similar size were built and imaged on a SOMATOM Flash scanner (Siemens Healthcare). Images were reconstructed using FBP and a Sinogram Affirmed Iterative Reconstruction (SAFIRE). Using an image subtraction technique, quantum noise was measured in all images of each phantom. It was found that in FBP, the noise was independent of the background (textured vs uniform). However, for SAFIRE, noise increased by up to 44% in the textured phantoms compared to the uniform phantom. As a result, the noise reduction from SAFIRE was found to be up to 66% in the uniform phantom but as low as 29% in the textured phantoms. Based on this result, it clear that further investigation was needed into to understand the impact that background texture has on image quality when iterative reconstruction algorithms are used.

To further investigate this phenomenon with more realistic textures, two anthropomorphic textured phantoms were designed to mimic lung vasculature and fatty soft tissue texture. The phantoms (along with a corresponding uniform phantom) were fabricated with a multi-material 3D printer and imaged on the SOMATOM Flash scanner. Scans were repeated a total of 50 times in order to get ensemble statistics of the noise. A novel method of estimating the noise power spectrum (NPS) from irregularly shaped ROIs was developed. It was found that SAFIRE images had highly locally non-stationary noise patterns with pixels near edges having higher noise than pixels in more uniform regions. Compared to FBP, SAFIRE images had 60% less noise on average in uniform regions for edge pixels, noise was between 20% higher and 40% lower. The noise texture (i.e., NPS) was also highly dependent on the background texture for SAFIRE. Therefore, it was concluded that quantum noise properties in the uniform phantoms are not representative of those in patients for iterative reconstruction algorithms and texture should be considered when assessing image quality of iterative algorithms.

The move beyond just assessing noise properties in textured phantoms towards assessing detectability, a series of new phantoms were designed specifically to measure low-contrast detectability in the presence of background texture. The textures used were optimized to match the texture in the liver regions actual patient CT images using a genetic algorithm. The so called “Clustured Lumpy Background” texture synthesis framework was used to generate the modeled texture. Three textured phantoms and a corresponding uniform phantom were fabricated with a multi-material 3D printer and imaged on the SOMATOM Flash scanner. Images were reconstructed with FBP and SAFIRE and analyzed using a multi-slice channelized Hotelling observer to measure detectability and the dose reduction potential of SAFIRE based on the uniform and textured phantoms. It was found that at the same dose, the improvement in detectability from SAFIRE (compared to FBP) was higher when measured in a uniform phantom compared to textured phantoms.

The final trajectory of this project aimed at developing methods to mathematically model lesions, as a means to help assess image quality directly from patient images. The mathematical modeling framework is first presented. The models describe a lesion’s morphology in terms of size, shape, contrast, and edge profile as an analytical equation. The models can be voxelized and inserted into patient images to create so-called “hybrid” images. These hybrid images can then be used to assess detectability or estimability with the advantage that the ground truth of the lesion morphology and location is known exactly. Based on this framework, a series of liver lesions, lung nodules, and kidney stones were modeled based on images of real lesions. The lesion models were virtually inserted into patient images to create a database of hybrid images to go along with the original database of real lesion images. ROI images from each database were assessed by radiologists in a blinded fashion to determine the realism of the hybrid images. It was found that the radiologists could not readily distinguish between real and virtual lesion images (area under the ROC curve was 0.55). This study provided evidence that the proposed mathematical lesion modeling framework could produce reasonably realistic lesion images.

Based on that result, two studies were conducted which demonstrated the utility of the lesion models. The first study used the modeling framework as a measurement tool to determine how dose and reconstruction algorithm affected the quantitative analysis of liver lesions, lung nodules, and renal stones in terms of their size, shape, attenuation, edge profile, and texture features. The same database of real lesion images used in the previous study was used for this study. That database contained images of the same patient at 2 dose levels (50% and 100%) along with 3 reconstruction algorithms from a GE 750HD CT system (GE Healthcare). The algorithms in question were FBP, Adaptive Statistical Iterative Reconstruction (ASiR), and Model-Based Iterative Reconstruction (MBIR). A total of 23 quantitative features were extracted from the lesions under each condition. It was found that both dose and reconstruction algorithm had a statistically significant effect on the feature measurements. In particular, radiation dose affected five, three, and four of the 23 features (related to lesion size, conspicuity, and pixel-value distribution) for liver lesions, lung nodules, and renal stones, respectively. MBIR significantly affected 9, 11, and 15 of the 23 features (including size, attenuation, and texture features) for liver lesions, lung nodules, and renal stones, respectively. Lesion texture was not significantly affected by radiation dose.

The second study demonstrating the utility of the lesion modeling framework focused on assessing detectability of very low-contrast liver lesions in abdominal imaging. Specifically, detectability was assessed as a function of dose and reconstruction algorithm. As part of a parallel clinical trial, images from 21 patients were collected at 6 dose levels per patient on a SOMATOM Flash scanner. Subtle liver lesion models (contrast = -15 HU) were inserted into the raw projection data from the patient scans. The projections were then reconstructed with FBP and SAFIRE (strength 5). Also, lesion-less images were reconstructed. Noise, contrast, CNR, and detectability index of an observer model (non-prewhitening matched filter) were assessed. It was found that SAFIRE reduced noise by 52%, reduced contrast by 12%, increased CNR by 87%. and increased detectability index by 65% compared to FBP. Further, a 2AFC human perception experiment was performed to assess the dose reduction potential of SAFIRE, which was found to be 22% compared to the standard of care dose.

In conclusion, this dissertation provides to the scientific community a series of new methodologies, phantoms, analysis techniques, and modeling tools that can be used to rigorously assess image quality from modern CT systems. Specifically, methods to properly evaluate iterative reconstruction have been developed and are expected to aid in the safe clinical implementation of dose reduction technologies.

Radiation therapy (RT) has been a standard-of-care treatment for many localized cancers for decades. Despite being an effective treatment modality for many clinical presentations, the efficacy of RT against cancer can be limited due to local recurrence, metastatic spread, and radiation resistance from tumor hypoxia. These limitations provide opportunity for innovative approaches to enhance the overall efficacy of RT. This thesis explores the potential novel approach to enhancing RT through the paradigm changing approach of adding a phototherapeutic component initiated simultaneously with RT. X-ray Psoralen Activated Cancer Therapy (X-PACT) is one such approach, where diagnostics-energy kilovoltage (kV) x-ray coupled with energy modulators (phosphors) converts kV photon to ultraviolet (UV) light, which in turn activates psoralen. Radiotherapy Enhanced with Cherenkov photo-Activation (RECA) is another approach, where therapeutic megavoltage (MV) x-ray generates UV light via Cherenkov phenomenon. Both approaches could increase local control in RT, increase treatment effectiveness in hypoxic tumors, and amplify anti-cancer systemic response. The overarching hypothesis that drives this dissertation is that X-PACT and RECA can activate psoralen to enhance cytotoxicity in-vitro and tumor growth control in-vivo compared to RT alone. In line with this hypothesis, this work explores the feasibility of both X-PACT and RECA via in-vitro and in-vivo verification as well as optimization of radiation techniques to maximize the therapeutic benefit of the approach.

X-PACT and RECA in-vitro / in-vivo studies indicate radiotherapy enhancement is plausible with psoralens activated by secondary UV light production from radiation, though further investigation is required to establish feasibility of RECA in-vivo. For X-PACT in-vitro, a substantial reduction in cell viability and increase in apoptosis was observed in various murine cancer cells (4T1, KP-B, and CT2A) when treated with a combination of 50µg/mL phosphor, 10µM psoralen (8-MOP), and 1Gy of 80kVp x-ray (viability < 20%), compared to any of these components alone (viability > 70%). This suggests a synergistic interaction between the components congruent with the X-PACT scheme, where x-ray induces phosphor UV emission, which in turn activates psoralen. The X-PACT in-vivo mice study showed improved survival with X-PACT versus saline control with flank 4T1 tumors (30.7 days for X-PACT vs. 21.6 days for saline) for survival criteria of 1000, 1500, and 2000mm3, respectively. For RECA, in-vitro results seem promising, where reductions in viability of 20% and 9.5% were observed for 4T1 and B16 murine cancer cell lines treated with RECA (radiation + trioxsalen, a potent psoralen derivative) versus radiation alone. A substantial increase in MHC I expression was observed for B16 cells treated with RECA versus those treated with radiation alone. A small RECA in-vivo pilot study using 8-MOP was inconclusive. Further in-vivo trials with a greater number mice per arm of are required to establish the RECA feasibility to enhance radiotherapy.

Feasibility of treatment optimization for both X-PACT and RECA were demonstrated with kV and MV beams respectively, by optimization of optical output per radiation dose delivered. It was found that in both X-PACT and RECA scheme, the energy of the photon radiation beam (i.e. tube voltage and LINAC energy settings) affected optical output the most. With kV beams for X-PACT, accurate beam delivery within the target volume to reduce normal tissue damage typically expected of kV beams was demonstrated with a 3D-printing-based preclinical irradiation scheme, which is expected to help X-PACT translation into the clinics. In addition, for X-PACT, novel MV-responding phosphors were characterized under MV radiation beam, suggesting the possibility of MV-radiation-mediated X-PACT. Immediate future studies should investigate the efficacy of the optimized X-PACT and RECA, as well as MV X-PACT in-vitro and in-vivo. Studies beyond these immediate ones should investigate X-PACT and RECA efficacy against hypoxic and metastatic tumor sites, where radiation can traditionally fail.

Background: Radiomics is a process of converting medical images into high-dimensional quantitative features and the subsequent mining these features for providing decision support. It is conducted as a potential noninvasive, low-cost, and patient-specific routine clinical tool. Building a predictive model which is reliable, efficient, and accurate is a vital part for the success of radiomics. Machine learning method is a powerful tool to achieve this. Feature extraction strongly affects the performance. Delta-feature is one way of feature extraction methods to reflect the spatial variation in tumor phenotype, hence it could provide better treatment-specific assessment.

Purpose: To compare the performance of using pre-treatment features and delta-features for assessing the brain radiosurgery treatment response, and to investigate the performance of different combinations of machine learning methods for feature selection and for feature classification.

Materials and Methods: A cohort of 12 patients with brain treated by radiosurgery was included in this research. The pre-treatment, one-week post-treatment, and two-month post-treatment T1 and T2 FLAIR MR images were acquired. 61 radiomic features were extracted from the gross tumor volume (GTV) of each image. The delta-features from pre-treatment to two post-treatment time points were calculated. With leave-one-out sampling, pre-treatment features and the two sets of delta-features were separately input into a univariate Cox regression model and a machine learning model (L1-regularized logistic regression [L1-LR], random forest [RF] or neural network [NN]) for feature selection. Then a machine learning method (L1-LR, L2-regularized logistic regression [L2-LR], RF, NN, kernel support vector machine [Kernel-SVM], linear-SVM, or naïve bayes [NB]) was used to build a classification model to predict overall survival. The performance of each model combination and feature type was estimated by the area under receiver operating characteristic (ROC) curve (AUC).

Results: The AUC of one-week delta-features was significantly higher than that of pre-treatment features (p-values < 0.0001) and two-month delta-features (p-value= 0.000). The model combinations of L1-LR for feature selection and RF for classification as well as RF for feature selection and NB for classification based on one-week delta-features presented the highest AUC values (both AUC=0.944).

Conclusions: This work potentially implied that the delta-features could be better in predicting treatment response than pre-treatment features, and the time point of computing the delta-features was a vital factor in assessment performance. Analyzing delta-features using a suitable machine learning approach is potentially a powerful tool for assessing treatment response.

Brain Positron emission tomography (PET) has been widely employed for the clinic diagnosis of Alzheimer's disease (AD). Studies have shown that PET imaging is helpful in differentiating healthy elderly individuals, mild cognitive impairment (MCI) individuals, and AD individuals (Nordberg, Rinne, Kadir, & Långström, 2010). However, PET image quality and quantitative accuracy is degraded from partial volume effects (PVEs), which are due to the poor spatial resolution of PET. As a result, the compensation of PVEs in PET may be of great significance in the improvement of early diagnosis of AD. There are many different approaches available to address PVEs including region-based methods and voxel-based methods. In this study, a voxel-based PVE compensation technique using high-resolution anatomical images was investigated. The high-resolution anatomical images could be computed tomography (CT) or magnetic resonance imaging (MRI) images. Such methods have been proposed and investigated in many studies (Vunckx et al., 2012). However, relatively little research has been done on comparing the effects of different MRI images on voxel-based PVE correction methods. In this study, we compare the effect of 6 different MRI image protocols on PVE compensation in PET images. The MRI protocols compared in this study are T1-, T2-, proton-density (PD)-weighted and 3 different inversion recovery MRI protocols.

Results: OSEM and MAP/ICD images with isotropic prior are blurry and/or noisy. Compared with the OSEM and MAP/ICD images obtained by using an isotropic prior, the PET image reconstructed using anatomical information show better contrast and less noise. Visually, the PET image reconstructed with the ZeroCSF prior gave the PET image that visually appears to match best with the PET phantom. PET images reconstructed with T2, PD and ZeroWM image are similar to one another in image quality, but relative to the PET phantom and the ZeroCSF PET image, these images have poor contrast between CSF pockets and surrounding GM tissue, and they have less contrast between GM and WM. PET image reconstructed with T1 image had a better GM and CSF contrast, some of the CSF pockets in GM were reconstructed, but the WM region was very noisy. PET images reconstructed with ZeroGM image had noticeably worse performance on the GM reconstruction. Analysis suggest that these effects are caused by differences in tissue contrast with different MRI protocols

Keywords: PET, MRI, partial volume effect, image reconstruction, SPECT, Alzheimer's disease.

Introduction: Medical imaging is essential in oncology for detecting, diagnosing, and treating cancer, and monitoring treatment effectiveness. Radiomics and machine learning are techniques that use computer algorithms to extract and analyze a vast number of quantitative features from medical images, which can lead to more accurate diagnoses and treatment plans. However, technical challenges, such as rare events and multi-object radiomics need to be addressed to fully realize the potential of these techniques in medical imaging and improve patient outcomes. Two examples of technical challenges in medical imaging are (1) the rare occurrence of a positive cancer diagnosis relative to the screened population, and (2) the difficulty of applying radiomics to multiple tumors in the same image, as seen in cases of multiple brain metastases.

Methods: (1) To evaluate the diagnostic performance of lung cancer screening (LCS) on low dose CT (LDCT), We retrospectively enrolled patients who received LCS via LDCT within our healthcare system between 1/1/2015-6/30/20. Our LCS program is a high-volume, ACR-recognized LCS program that houses a structured reporting registry of Lung-RADS scores. Using data from the electronic health record, we defined a malignant pulmonary nodule (i.e., lung cancer) as a pathology-proven diagnosis of lung cancer (via tissue obtained from a needle biopsy, bronchoscopy, or surgical biopsy). We determined the rate of screen-detected lung cancers, as well as all lung cancers diagnosed within one year after a LCS exam. The diagnostic performance of LCS was determined based on receiver operating characteristic analysis. Relevant clinical and demographic characteristics were analyzed as potential confounding factors, including age, sex, race/ethnicity, and smoking history. Predictive modeling on support vector machine (SVM) was performed and compared to standard-of-care Lung-RADS. (2) To explore radiomic feature aggregation methods in patients with metastatic brain cancer, seventy-eight relevant radiomic features were extracted from 449 unique metastases from 159 unique patients treated with stereotactic radiosurgery (SRS) using SPGR or T1+c MRI scans. MRI scans were normalized and discretized into 64 gray levels. Three different aggregation techniques were evaluated to compare radiomic feature results: (1) simple average, (2) weighted average by tumor volume, and (3) weighted average of the three largest metastases by volume. Univariate Kaplan-Meier analysis was performed based on the median value of each feature for three distinct clinical endpoints: overall survival, intracranial progression-free survival (ICPFS), and extracranial progression-free survival (ECPFS). In addition, this study considered molecular drivers (including EGFR, ALK, BRAF, KRAS, PD-L1, ROS1) and some clinical/demographic factors (age at SRS, KPS, number of metastases and NSCLC type) as potential confounding variables, evaluated for radiogenomic association based on Fisher's Exact Test.

Results: (1) 5,150 LCS exams were performed on 3,326 unique patients. The average age at LCS was 65.4±6.2 years, with 51.4% (1709/3,326) being male. The sensitivity and specificity of LCS were 93.1% and 83.8% respectively. Patients with positive Lung-RADs scores and patients who were current smokers had a higher likelihood of screen-detected lung cancer than former smokers (p<0.001 and p=0.017 respectively). The sensitivity plus specificity of one-class training on SVM outperformed standard-of-care Lung-RADS alone. (2) Radiomic texture features Small Zone Emphasis (p=0.014) and Correlation (p=0.018) demonstrated a significant association with ICPFS and ECPFS, respectively, regardless of the feature aggregation technique. The radiomic morphological feature Compactness was also significant for these endpoints, suggesting that both tumor shape and volume-corrected texture provide complementary prognostic value. The EGFR mutation was found to be associated with 11 prognostically-relevant radiomic features for ECPFS, with the strongest association for the feature of Correlation (p=0.010).Conclusions: (1) LCS has high sensitivity, modest specificity, and relatively low PPV, the latter suggesting a need for improvements in classification of "positive" LCS results. Screen-detected lung cancers were likely in currently smoking patients. (2) This exploratory study identified several associations between radiomic features and clinical endpoints, providing insight into their potential prognostic value. Molecular drivers were also identified as confounding variables, emphasizing the importance of further radiogenomic analyses in brain metastases.

Our group has shown X-ray diffraction imaging for thin samples, however, its applicability to thick samples for pathology diagnostics, small animal imaging, and potentially in-vivo applications has yet to be explored. Single scatter events dominate in tissues on the order of a few millimeters, and solving the inverse problem of scatter localization is straightforward. As the sample thickness increases, multiple scatter and geometric blurring effects become important. We look to quantify the role of optical and geometric object thickness, explore the effects of tumor size and location, and quantify the role of varying breast density in medical X-ray diffraction imaging. Using MC-GPU, a Monte Carlo GPU-accelerated photon simulation code, we simulate X-ray diffraction studies for various energies with phantoms combining a variety of object compositions, tumor sizes, tumor locations, and distances from the detector. Previous validation of MC-GPU conducted by our group has added additional form factors allowing for the implementation of the molecular interference function. A notional pencil beam diffraction system is modeled with a 2D grid of pixels approximating a 2D flat panel, noise-free, and 100% efficient detector. Results are analyzed using different metrics, including signal-to-background, signal-to-noise, and signal-to-multiple scatter. Analysis showed optimal energy ranges for thick tissue diffraction simulations, non-significant effects of tumor location in the object, and varying breast composition playing a pivotal role in tumor detectability. This work has indicated the potential for significant advances in medical X-ray imaging, specifically in the application to in vivo and thick tissue imaging. We provide evidence that X-ray diffraction imaging on thick tissue samples is feasible under proper conditions. Overall, we further our understanding of the role of thick tissues, tumor location, and tumor size in medical X-ray diffraction imaging and provide a framework for analyzing and implementing XRD imaging on thick samples.