Binary and Multi-Class Classification of Colorectal Polyps Using CRP-ViT: A Comparative Study Between CNNs and QNNs.

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Oncology
• Quantum Computing Applications

Background:

• Colorectal cancer (CRC) poses a significant global health burden, with polyps as key precursors.
• Accurate classification of colorectal polyps (CRPs) from colonoscopy images is crucial for early CRC diagnosis and treatment.

Purpose of the Study:

• To develop and evaluate a novel hybrid model, CRP-ViT, for enhanced colorectal polyp classification.
• To compare the performance of CRP-ViT against traditional CNNs and emerging QNNs.

Main Methods:

• Proposed a hybrid CRP-ViT model combining ResNet50 and Vision Transformers (ViTs) for feature extraction.
• Conducted binary and multi-class classification experiments for polyp detection and type prediction.
• Compared CRP-ViT performance with CNNs and QNNs, focusing on accuracy and computational efficiency.

Main Results:

• The CRP_QNN-ViT model demonstrated superior classification performance in both binary and multi-classification tasks.
• Achieved high accuracy rates: 98.18% (train) and 97.73% (validation) for binary classification; 98.13% (train) and 97.92% (validation) for multi-classification.
• CRP_QNN-ViT exhibited enhanced computational efficiency, particularly in terms of processing time.

Conclusions:

• The integration of quantum computing shows promise for advancing medical image analysis.
• Transformer-based architectures, like ViTs, are highly effective for classifying colorectal polyps.