A multi-task convolutional neural network for classification and segmentation of chronic venous disorders

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Vascular Medicine

Background:

• Chronic Venous Disorders (CVD) are highly prevalent, affecting 35% of adults and posing increasing healthcare challenges due to aging populations.
• Current CVD diagnosis relies on subjective visual assessment, which is time-consuming and expertise-dependent.
• Accurate diagnosis and monitoring are crucial for effective CVD treatment planning and patient management.

Purpose of the Study:

• To develop an automated strategy for the joint segmentation and classification of Chronic Venous Disorders (CVD).
• To leverage a multi-task deep learning network (VENet) for improved efficiency and performance in CVD analysis.
• To enhance the robustness and accuracy of CVD diagnosis and lesion segmentation.

Main Methods:

• A novel multi-task deep learning network, VENet, was designed to perform simultaneous segmentation and classification of CVDs.
• The VENet model was trained and evaluated on a dataset comprising 1376 CVD images.
• Performance was benchmarked against existing state-of-the-art methods for both classification and segmentation tasks.

Main Results:

• VENet achieved high classification performance: 96.4% accuracy, 96.4% precision, and 97.2% recall.
• Segmentation performance metrics included a 75.4% Dice coefficient, 76.7% precision, and 76.7% recall.
• The joint learning approach demonstrated increased robustness for both segmentation and classification compared to standalone methods, particularly for small lesions.

Conclusions:

• The proposed VENet model offers an effective automated solution for the joint segmentation and classification of CVDs.
• This AI-driven approach shows significant potential to improve the accuracy and efficiency of diagnosing and monitoring lower limb venous disorders.
• The multi-task learning strategy enhances diagnostic capabilities, especially in identifying subtle or small venous lesions.