Improving Prostate Cancer Detection using Multiparametric Ultrasound

Prostate cancer (PCa) is the second most common cancer diagnosis, behind skin cancer, and the second most common cause of cancer-related death, behind lung cancer, for men in the United States. The prevalence of PCa increases with age and ranges from 1.8% of men being diagnosed with PCa before age 59 to 11.6% of men being diagnosed with PCa over the course of their entire lives. PCa is typically diagnosed using transrectal ultrasound (TRUS) guided biopsy which commonly consists of 10-12 systematically sampled biopsy cores taken from specified regions within the prostate. In TRUS guided biopsy, the TRUS B-mode imaging is used by the clinician to ensure the biopsy needles remain within the prostate but is not sensitive nor specific enough to identify and target PCa-suspicious regions. Multiparametric magnetic resonance imaging (mpMRI) fusion biopsy is the current gold standard for targeted PCa biopsy, though this approach comes at added cost and is not widely available. mpMRI fusion biopsy also requires the registration of the pre-biopsy mpMRI with real-time TRUS B-mode imaging which can result in an incorrectly targeted lesion due to registration error.This thesis explores advanced ultrasound techniques, such as acoustic radiation force impulse (ARFI) imaging, shear wave elasticity imaging (SWEI), quantitative ultrasound’s (QUS) midband fit parameter (MF), and multiparametric ultrasound (mpUS), for PCa identification and targeting during biopsy. The goals of this thesis are to (1) establish a shear wave speed (SWS) threshold for identifying PCa using SWEI, (2) create an mpUS approach which combines ARFI, SWEI, MF, and B-mode imaging and assess the improvement in PCa visibility when using mpUS and (3) assess the performance of ARFI, SWEI, MF, and mpUS when locating suspicious regions which align with mpMRI-identified PCa to provide registration validation during fusion biopsy. Combined, this thesis provides preliminary data and motivation for future work developing and assessing advanced ultrasound imaging methods for image-guided targeted prostate biopsy. The data included throughout this thesis was acquired using a custom ultrasound setup capable of acquiring both elasticity (ARFI and SWEI) and acoustic backscatter (B-mode and MF) data in a single imaging session. Additionally, the ultrasound system was paired with a rotation stage allowing for 3D data acquisition which yielded co-registered image volumes for each of the four ultrasound modalities. This data was acquired in patients immediately preceding radical prostatectomy. Histopathology analysis of the excised prostates was used to determine the ground truth locations of PCa for each patient, allowing for the labeling of the ultrasound data as PCa or healthy tissue. In Chapter 3, the SWEI data volumes were used to identify a shear wave speed (SWS) value threshold to separate PCa from healthy prostate tissue. This SWS threshold yielded sensitivities and specificities akin to mpMRI fusion biopsy. Additionally, a SWS ratio was assessed to normalize for tissue compression and patient variability. This threshold was accompanied by a substantial increase in specificity, positive predictive value (PPV), and area under the receiver operating characteristic curve (AUC). This section demonstrates the feasibility of using 3D SWEI data to detect and localize PCa and demonstrates the benefits of normalizing for applied compression during data acquisition for use in biopsy targeting studies. In Chapter 4, a linear support vector machine (SVM) was used to combine B-mode imaging, ARFI, SWEI, and MF into a synthesized mpUS volume to enhance lesion visibility. mpUS led to improvements in lesion visibility metrics compared to each individual ultrasound modality. The individual advanced ultrasound modalities (ARFI, MF, and SWEI) also all outperformed B-mode in contrast. The improved performance of mpUS demonstrates the benefit of combining ultrasound techniques based on different contrast mechanisms, supporting its utility for ultrasound-based targeted prostate biopsy. In Chapter 5, the histologically determined PCa locations were identified in mpMRI’s T2 and apparent diffusion coefficient (ADC) images and compared to the corresponding regions in B-mode, ARFI, SWEI, MF, and mpUS images. SWEI only failed to identify one PCa lesion in the posterior of the prostate and B-mode, MF, and ARFI all successfully identified 100% of the anterior lesions, indicating that, when combined, advanced ultrasound techniques facilitate the visualization of the majority of mpMRI-identified regions of interest throughout the entire prostate. Additionally, the mpUS combination developed in Chapter 4 was applied to a subset of 10 patients and resulted in correct localization of 88% (14/16) of the mpMRI-identified lesions. This work demonstrates the feasibility of using advanced ultrasound techniques to locate mpMRI-identified lesions, which would enable improved registration validation during fusion biopsy. Finally, Chapter 6 includes further insights into this work and the implications it may have on the diagnosis of PCa. Advanced ultrasound is a promising approach for both targeted PCa biopsy and for screening. Additionally, combining information from multiple advanced ultrasound techniques (ARFI, SWEI, and MF) yields improved performance over any single method indicating that the future of prostate ultrasound is multiparametric.