The proposed system presents a multi-task framework for automated lung infection analysis using high-resolution computed tomography (HRCT) images. It integrates two novel deep learning architectures: MS-CSTA-Net for multi-class classification of lung infections and DualMask-SegNet v5.4 for precise segmentation of affected regions. The classification module employs a multi-scale cross-scale token attention mechanism to capture both global and local features, enhancing recognition of subtle infection patterns. The segmentation module leverages dual mask learning to accurately delineate infection boundaries, improving interpretability and feature extraction for downstream classification tasks. The system is built on a Flask-based framework, offering a web interface with modules for Home, Register, Login, Prediction or Classification, and Logout. Users can input HRCT images, and the system provides segmentation maps alongside predicted infection classes. Experimental evaluation demonstrates improved performance in terms of accuracy, precision, recall, and F1-score, highlighting the model’s capability to handle complex pulmonary image patterns. The framework offers an efficient and scalable solution for lung infection analysis, combining robust feature extraction, attention mechanisms, and dual-task learning in a single unified system.

Keywords: HRCT images, lung infection, segmentation, classification, MS-CSTA-Net, DualMask-SegNet, attention mechanism, deep learning, feature extraction, multi-task learning.