A review of deep learning-based multiple-lesion recognition from medical images: classification, detection and...

Area of Science:

• Medical Image Analysis
• Artificial Intelligence in Medicine
• Computer Vision

Background:

• Deep learning (DL) is dominant in medical image processing, excelling at single lesion recognition and segmentation.
• Multiple-lesion recognition presents significant challenges due to subtle variations and wide lesion ranges.
• Recent advancements have spurred the development of DL algorithms to address these complexities.

Purpose of the Study:

• To provide a comprehensive overview and critical analysis of recent deep learning-based methods for multiple-lesion recognition.
• To examine the application of these methods in diverse body areas and for whole-body multiple disease recognition.
• To identify persistent challenges and suggest future research directions in the field.

Main Methods:

• Review and analysis of recent deep learning-based algorithms for multiple-lesion recognition.
• Assessment of methods applied to diverse anatomical regions and multi-disease scenarios.
• Critical evaluation of existing literature and identified challenges.

Main Results:

• Deep learning methods show promise but face difficulties in handling variations and ranges in multiple lesions.
• Current approaches are being applied to various body parts and complex multi-disease recognition tasks.
• Significant challenges persist, indicating a need for further methodological development.

Conclusions:

• Deep learning is a powerful tool for medical image analysis, but multiple-lesion recognition requires further innovation.
• Addressing variations between lesions and the broad spectrum of diseases is crucial for future advancements.
• This review highlights key challenges and future research avenues to guide the development of more robust algorithms.