ABSTRACT
The goal of this chapter is to build a computer-aided diagnosis (CAD) system for diagnosing prostate cancer from diffusion-weighted magnetic resonance imaging (DWI). The first step in the proposed system segments the prostate using a level-set model. This model is guided by a stochastic speed function that is derived using non-negative matrix factorization (NMF). The NMF attributes are calculated by fusing prostate voxels’ intensities, probabilistic shape model, and the spatial interactions between prostate voxels. In the second step, the apparent diffusion coefficient (ADC) of the segmented prostate volume is mathematically calculated at distinct b-values. To preserve continuity, the calculated ADC values are normalized and refined using a Generalized Gauss–Markov Random Field (GGMRF) image model. The cumulative distribution function (CDF) of refined ADC for the prostate tissues at distinct-values are then constructed. These CDFs are considered as global features which can be used to distinguish between benign and malignant tumors. A K-nearest-neighbor is fed with the blood test results to produce blood test-based probabilities for the different cases. In the final step, a two-stage structure of stacked non-negativity constraint auto-encoder (SNCAE) is trained to classify the prostate tumor as benign or malignant based on the constructed CDFs. In the first stage, classification probabilities are estimated at each b-value, and in the second stage, those probabilities together with the blood test-based probabilities are fused and fed into the prediction stage SNCAE to calculate the final classification. According the preliminary experiments on 18 DWI datasets, the system accuracy improved by fusing the blood test results with DWI results.
Keywords: Computer-aided diagnosis, Autoencoders, DWI.
