Abstract:
Prostate Cancer (PCa) is one of the most common type of cancer found in men aged over 45 years. Detection and staging is the most critical step for pathologist and treatment decision is based on the findings. PCa is graded according to the Gleason grading criteria which is based on the glandular and pattern of the cells in the histological digitized biopsy slides. Manual grading depends on the experience of pathologists, quality of staining and some other parameters. To reduce these issues, an automatic grading system is required and this research will support the development of accurate computer aided design (CAD) system for the automatic detection and grading of prostate cancer. The goal of this study is to differentiate between benign and malignant tissues, classify the H&E stained histological images into one of the 3 grades as per Gleason grading criteria. Prostatic images data collection is the most challenging task in this research which took around 4 months. After visiting Shaukat Khanum Hospital several times, collected the H&E stained digitized images collected from around 150 patients. Dataset has been divided into four categories including grade 1-2, grade 3, grade 4 and grade 5 having 56, 69, 74 and 69 images respectively. In this research, we have extracted several features set using Gabor and local binary patterns (LBP) with its variants and combined the features to enhance the performance. We have applied multiple classifiers including K-Nearest Neighbours (KNN), Support Vector Machine (SVM) and Random Forest for evaluating performance. After analysing performance, we developed new features space by combining Gabor energy and rotation invariant LBP descriptors. It shows highest accuracy while having very low and simple feature dimensions. The proposed system shows overall average accuracy of 98.26 % while grading benign, grade 3, 4 & 5 grades.
