DeepCardioNet: A Temporally Aware, Attention-Guided Multi-Task Network for Joint Cardiac Structure Segmentation and Disease Diagnosis from Cine Cardiac MRI

DeepCardioNet is a temporally aware, attention-guided multi-task deep learning framework for simultaneous cardiac structure segmentation and disease diagnosis using cine cardiac MRI. By integrating ResNet-18, bidirectional ConvLSTM, and joint channel–spatial attention, the model captures spatial and temporal cardiac features to improve diagnostic accuracy and segmentation performance. Evaluated on the ACDC and MIMIC-CXR datasets, the proposed approach demonstrates promising results for efficient, end-to-end automated cardiac image analysis, with potential applications in clinical decision support.