Deeply Supervised Active Learning for Finger Bones Segmentation.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Vision

Background:

• Medical image segmentation is crucial but challenging.
• Accurate segmentation of finger bones is vital for clinical diagnosis and treatment planning.
• Current methods often require extensive labeled data, increasing costs and time.

Purpose of the Study:

• To develop a novel deeply supervised active learning approach for efficient finger bone segmentation.
• To reduce the need for large annotated datasets in medical image analysis.
• To improve the practicality and generalizability of automated segmentation techniques.

Main Methods:

• A novel deeply supervised active learning framework was proposed.
• The architecture was fine-tuned iteratively and incrementally.
• A deep supervision mechanism guided hidden layer learning and sample selection for labeling.

Main Results:

• The method achieved competitive segmentation results for finger bones.
• Significantly fewer labeled samples were required compared to full annotation.
• The approach demonstrated effectiveness in reducing annotation burden.

Conclusions:

• The proposed method offers an efficient solution for finger bone segmentation using limited annotations.
• This approach has potential clinical relevance for medical practices and can be generalized to other applications.
• Deeply supervised active learning can significantly improve the efficiency of medical image segmentation tasks.