Medlightnet: A Lightweight Multi-Colour-Space CNN for Childhood Medulloblastoma Diagnosis From Histopathological Images

Early and accurate diagnosis of Childhood MedulloBlastoma (CMB) is essential for timely intervention and improved prognosis. Histopathological image analysis has greatly benefited from the integration of artificial intelligence. However, conventional models often rely solely on the RGB colour space, limiting the ability to capture diverse visual cues present in tumour tissues. This paper introduces a novel Multi-Colour-Space Decision Support System (MCSDSS) for CMB prediction using histopathological images, incorporating RGB, HSV, and LAB colour spaces to enhance feature representation. A custom-designed lightweight Convolutional Neural Network (CNN), MedLightNet, is designed to extract deep features from each colour space individually. The extracted features are fused to generate a comprehensive diagnostic biomarker for CMB prediction. Experimental validation demonstrates that the proposed MCSDSS powered by MedLightNet, achieves superior diagnostic performance compared to traditional single-space and conventional deep learning models, with high accuracy, sensitivity, and specificity.