الفهرس 
Chapter 1: IntroductionOnly 14 pages are availabe for public view
 |  | from | 152 |  |  |
| | | from | 152 | | |
Abstract
Early diagnosis of prostate cancer increases the chances of patients’ survival. The proposed automated non-invasive computer-aided diagnosis (CAD) of prostate cancer performs three main steps. It segments the prostate on diffusion-weighted magnetic resonance images (DW-MRI) acquired at seven different b-values, estimates its apparent diffusion coefficient (ADC), and classifies their descriptors – empirical cumulative distribution functions (CDFs) – with a trained deep learning network. To segment the prostate, an evolving geometric (level-set-based) deformable model is guided by a speed function depending on intensity attributes extracted from the DW-MRI. For a more robust evolution, the attributes are fused with a probabilistic shape prior and estimated spatial dependencies between prostate voxels using nonnegative matrix factorization (NMF). The ADCs of the segmented prostate volume at different b-values are estimated as discriminatory features. The CDFs of the refined ADCs at different b-values are considered global water diffusion features and used to distinguish between benign and malignant prostates. A deep learning network of stacked non-negativity-constrained auto-encoders (SNCAE) is trained to classify the benign or malignant prostates based on the constructed CDFs. Different experiments were conducted to evaluate the different structures proposed for diagnosing prostate cancer. The first experiment integrated imaging markers with clinical bio-markers of 18 subjects. The accuracy increased from 88.9%, when using only the imaging markers, to 94.4% after integrating the imaging markers with clinical bio-markers. Then, a series of experiments were conducted to evaluate two different structures of SNCAE using a larger number of subjects (53 DW-MRI data sets). These structures are single-phase structure of SNCAE and two-phase structure of SNCAE. The highest accuracy of the single-phase resulted at b-value = 700 s/mm2. After integrating the imaging markers of all the seven different b-values using the two-phase structure of SNCAE, the resulting accuracy outperformed the accuracy of the single-phase structure at each individual b-value. The overall accuracy, sensitivity, and specificity after the second phase of the two-phase structure of SNCAE are 98.1%, 96.2%, and 100%, respectively. The conducted experiments show the importance of integrating the prostate specific antigen (PSA) screening results with the imaging markers. This integration increases the diagnosis accuracy, especially in the case of small data sets. In the case of larger data sets, the accuracy of diagnosis is improved by using DW-MRI data sets acquired at lower and higher b-values.