Deep learning-based object detection algorithms in medical imaging: Systematic review

Area of Science:

• Computer Science
• Medical Imaging
• Artificial Intelligence

Background:

• Deep Learning (DL) has significantly advanced various fields, particularly in medical image analysis for tasks like segmentation, detection, and classification.
• DL-based object recognition in medical imaging is gaining traction due to its potential to improve diagnostic accuracy and efficiency.

Purpose of the Study:

• To provide an overview of Deep Learning-based object recognition in medical images.
• To explore recent methods, imaging techniques, and anatomical applications of DL object detection.
• To analyze trends and identify future research directions in this domain.

Main Methods:

• A systematic literature review using PRISMA guidelines.
• Quantitative and qualitative analysis of publications based on citation rates.
• Examination of DL object detection utilization across different imaging modalities and anatomical domains.

Main Results:

• A consistent increase in the use of DL-based object detection models in medical imaging was observed.
• Research is most active in the US, China, and Japan, primarily within Medicine and Computer Science.
• DL methods are adaptable, applied to CR scans, pathology images, and endoscopic imaging, with notable use in digital pathology and microscopy.

Conclusions:

• Deep learning object detection shows significant, yet unexploited, potential in medical image analysis.
• Varied dataset sizes, a prevalence of private datasets, and a scarcity of prospective studies present challenges.
• Ongoing research and application-specific optimization are crucial for advancing DL in medical imaging.