Deep-Learning Oral-Lesion Classification with Multi-Annotator Bounding-Box Fusion
An applied medical-imaging approach that detects and classifies oral-cavity lesions from clinical photographs for low-cost early oral-cancer screening. It pairs a Deep Belief Network with a Capsule Network (CapsNet) to classify images as lesion-present/absent and referral/no-referral, and separately detects (localises) lesions for referral. Its distinguishing methodological contribution is an annotation-fusion procedure that reconciles bounding-box labels drawn independently by several clinical experts into one consolidated ground-truth label, addressing inter-annotator disagreement in the training set.
ENHANCED CLASSIFICATION OF ORAL CANCER USING DEEP LEARNING TECHNIQUES Dr. Senthil Pandi S Associate
A deep-learning study for automated detection and classification of oral-cavity lesions from clinical photographs, aimed at low-cost early screening for oral cancer in low- and middle-income regions.…