A survey on deep learning in medical image analysis.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Vision

Background:

• Deep learning algorithms, particularly convolutional networks (CNNs), are increasingly prevalent in medical image analysis.
• The rapid advancement of deep learning has led to significant contributions in the field over the past year.

Purpose of the Study:

• To review fundamental deep learning concepts relevant to medical image analysis.
• To summarize and categorize over 300 recent contributions to deep learning in medical imaging.
• To provide an overview of deep learning applications across diverse medical specialties.

Main Methods:

• Literature review of over 300 studies on deep learning in medical imaging.
• Categorization of studies by task: image classification, object detection, segmentation, and registration.
• Survey of applications in neuroimaging, retinal imaging, pulmonary imaging, digital pathology, breast imaging, cardiac imaging, abdominal imaging, and musculoskeletal imaging.

Main Results:

• Deep learning is effectively applied to various medical image analysis tasks, including classification, detection, and segmentation.
• Significant progress has been made across multiple medical application areas, with a high volume of research published recently.
• The review consolidates a broad spectrum of deep learning methodologies and their impact on medical diagnostics and research.

Conclusions:

• Deep learning, particularly CNNs, represents a state-of-the-art methodology for medical image analysis.
• Open challenges remain, highlighting critical areas for future research and development in the field.
• The continued evolution of deep learning promises further advancements in medical image interpretation and clinical applications.