An Ensemble Deep Transfer Learning Model for Multi-dimensional Image Classification of Histological Prostate Biopsy Patterns
Md Hafizur Rahman, Subrata Bhattacharjee, Yeong-Byn Hwang, Hee‐Cheol Kim, Heung‐Kook Choi · Journal of Korea Multimedia Society · 2024
Early prostate cancer diagnosis by pathologists remains challenging. Recent advances in computer-aided detection (CAD), artificial intelligence (AI), and machine learning (ML) allow prostate cancer grading. This study explored the accuracy of prostate cancer detection by deep learning techniques, particularly convolutional neural networks (CNNs). We performed three-way binary classification based on images cropped to 256 × 256 and 512 × 512 pixels using an ensemble deep CNN model. Six pre-trained CNN models (MobileNet, VGG-16, ResNet-50, DenseNet-121, Inception-V3, and EfficientNet-B0) were integrated to classify histopathological features. The overall accuracy for the combined 256 × 256 and 512×512 pixel images was 94.9%. Additionally, in separate classifications of 256×256 and 512×512 images, we achieved overall accuracies of 90.8% and 94.3%, respectively. Consequently, our method effectively distinguishes benign from malignant samples, approaching near-perfect accuracy.