About this Research Topic
The advances of the recent decade in Deep Learning had a significant impact on biomedical image segmentation and analysis. The state-of-the-art in several domains has been conquered by methods based on Convolutional Neural Networks (CNNs), such as U-Net, with comparable performance to what is obtained by experts.
Still, several issues complicate the successful adoption of such methods in clinical practice:
1) the parameterization of deep architectures requires effort and expertise, whereas the accuracy obtained is often sensitive to small variations
2) deep architectures often appear as ‘black boxes’ and their function cannot be intuitively interpreted by domain experts
3) learning is often targeted to specific imaging devices and a trained model is not directly transferable
4) annotating image regions to create labeled datasets is often hard
Deep Learning developments including Automated Parameterization, Explainable Learning, Transfer Learning or Weak Labeling/Annotation have not been fully adopted in the context of biomedical image segmentation and analysis.
Themes of interest in this Research Topic address biomedical image segmentation and analysis and include, but are not limited to, the following:
• Theories and models related to image segmentation and analysis with potential biomedical applications
• Segmentation, classification, time-series analysis, prediction, regression
• Hybrid methods combining rule-based and data-driven models
• Automated parameterization techniques to reduce the number or impact of hyper-parameters
• Explainable Learning
• Transfer Learning
• Weak Labeling/Annotation
• Active Learning
Article types which are welcome for submission include: original research, systematic review, methods, review, mini review, hypothesis and theory, perspective, data reports and opinion articles.
Still, several issues complicate the successful adoption of such methods in clinical practice:
1) the parameterization of deep architectures requires effort and expertise, whereas the accuracy obtained is often sensitive to small variations
2) deep architectures often appear as ‘black boxes’ and their function cannot be intuitively interpreted by domain experts
3) learning is often targeted to specific imaging devices and a trained model is not directly transferable
4) annotating image regions to create labeled datasets is often hard
Deep Learning developments including Automated Parameterization, Explainable Learning, Transfer Learning or Weak Labeling/Annotation have not been fully adopted in the context of biomedical image segmentation and analysis.
Themes of interest in this Research Topic address biomedical image segmentation and analysis and include, but are not limited to, the following:
• Theories and models related to image segmentation and analysis with potential biomedical applications
• Segmentation, classification, time-series analysis, prediction, regression
• Hybrid methods combining rule-based and data-driven models
• Automated parameterization techniques to reduce the number or impact of hyper-parameters
• Explainable Learning
• Transfer Learning
• Weak Labeling/Annotation
• Active Learning
Article types which are welcome for submission include: original research, systematic review, methods, review, mini review, hypothesis and theory, perspective, data reports and opinion articles.
Keywords: Image Segmentation, Image Analysis, Deep Learning, Convolutional Neural Networks, Recurrent Neural Networks, Neural Network Visualization, Biomedical Imaging, Automated Parameterization, Explainable Learning, Transfer Learning, Weak Labeling, Weak Annotat
Important Note: All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.
