Malignancy detection performance using excised breast tumor margin spectroscopic data and an optimal decision fusion based approach
Approximately 20-70% of women with breast cancer who choose to undergo breast-conserving surgery (BCS) need to return to the operating table for re-excision . Now devices utilizing optical spectroscopy are emerging as a new platform for intra-operative tumor margin assessment. This study aims to evaluate an optimal decision fusion approach for malignancy detection of measured spectroscopic data from a first-generation optical visible spectral imaging platform that can image the molecular composition of breast tumor margins by implementing a Monte Carlo method with measured diffuse reflectance [1, 2]. The device measures the diffuse reflectance across 450-600nm. After implementing the Monte Carlo algorithm the absorption and scattering spectra is derived and is used to provide insight on different optical properties present in the tissue mass . Although the extracted optical properties may provide insight on the biological composition of a specimen, it may not be ideal for malignancy detection. Demographic factors may also affect a women's normal tissue breast composition, which makes malignancy detection more complicated. This optimal decision fusion approach implements the basic decision fusion methodology on acquired spectroscopic data to evaluate the effect on malignancy detection for different extracted optical parameters. The results of this automated and systematic approach indicate that a performance of 90% sensitivity and 68% specificity can be achieved with this approach for the diffuse reflectance spectrum, which outperforms the extracted optical properties. However, when only considering post-menopausal patients, the absorption spectrum can yield a sensitivity of 90% and specificity of 82% and has the best performance of all other features for this demographic group.
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 United States License.
Rights for Collection: Masters Theses