Deep Learning-Based Multi-Class Model in WSIs: Vision Transformer for Distinguishing Cancer, Tumor, Mitosis, Red Blood Cells, and Karyorrhexis

Whole Slide Images (WSIs) contain complex cellular structures that demand advanced computational techniques for precise multi-class classification in digital pathology. This study presents a Deep Learning-Based Multi-Class Vision Transformer (ViT) model designed to accurately distinguish cancer cells, tumor regions, mitosis events, red blood cells (RBCs), and karyorrhexis in high-resolution WSIs. The proposed method integrates patch-based feature extraction with transformer attention mechanisms to effectively capture global contextual relationships across histopathological patterns. WSIs are divided into fixed-size patches, which are then embedded into a high-dimensional feature space using a patch embedding module. A customized Vision Transformer comprising multi-head self-attention layers and MLP expansion blocks is employed to learn discriminative representations of cellular morphology and tissue architecture. Extensive data augmentation enhances robustness against variations in staining, orientation, and cell size. A multi-layer perceptron (MLP) classification head is used to generate final predictions across multiple classes. Experimental evaluation demonstrates strong performance in terms of accuracy, precision, recall, F1-score, and Hamming loss, confirming the model's ability to differentiate subtle pathological features. The results highlight the effectiveness of transformer-based architectures for multi-class classification in WSIs, offering a reliable computational tool to support pathologists in early diagnosis and detailed tissue analysis.

Keywords: Whole Slide Images (WSIs), Vision Transformer (ViT), Multi-Class Classification, Histopathology Analysis, Deep Learning.