Deep learning has recently shown impressive performance in various health informatics applications. In particular, it has emerged as the de facto standard in many computational methods in digital pathology and has provided support for diagnosis, prognosis and treatment planning. However, deep neural networks typically require a large amount of data for model training, which is not always available due to privacy issues and/or budget deficits. This can significantly impede the deployment of state-of-the-art artificial intelligence (AI) or deep learning algorithms in real-world basic and clinical research as well as clinical practice. In addition, integration of multi-modal data, such as medical images, omics data, clinical notes, and lab test results, is under-explored in computational pathology, and many current methods do not take advantage of the rich information in different data modalities.
Although recent studies have introduced some AI approaches or applications aiming to address the issues of data shortage and/or heterogeneous data integration, there are still many unsolved problems and unexplored areas in computational pathology. In particular, digital pathology exhibits some unique challenges, such as high image dimension for whole-slide images, batch effects due to inconsistent imaging procedures, and often higher cost in data collection and annotation compared with other modalities of imaging data. These can worsen the situation where deep models are learned with limited data or integration of multi-modal data. By assembling a variety of viewpoints and a wealth of expertise, this Research Topic aims to offer a platform for computational pathology researchers to present cutting-edge methods and practical applications, and also for the field to promote collaborations among diverse groups.
This Research Topic invites AI experts, medical researchers, and physicians to contribute original research articles, reviews, and case studies, which focus on data/annotation-efficient modeling and integration of multi-modal data for computational pathology. The topics of interest include, but are not limited to:
- Weakly, active, semi- and self-supervised learning in digital pathology.
- Domain adaptation and generalization for cross-domain pathology data.
- Data synthesis including pathology image generation and translation.
- Federated learning to bridge pathology data silos.
- Multi-modal learning with different types of medical data.
- Geometric deep learning including graph neural networks for computational pathology.
Keywords:
Deep learning, Weakly-supervised learning, Self-supervised learning, Federated learning, Multi-modal learning, Generative models, Graph neural networks
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.
Deep learning has recently shown impressive performance in various health informatics applications. In particular, it has emerged as the de facto standard in many computational methods in digital pathology and has provided support for diagnosis, prognosis and treatment planning. However, deep neural networks typically require a large amount of data for model training, which is not always available due to privacy issues and/or budget deficits. This can significantly impede the deployment of state-of-the-art artificial intelligence (AI) or deep learning algorithms in real-world basic and clinical research as well as clinical practice. In addition, integration of multi-modal data, such as medical images, omics data, clinical notes, and lab test results, is under-explored in computational pathology, and many current methods do not take advantage of the rich information in different data modalities.
Although recent studies have introduced some AI approaches or applications aiming to address the issues of data shortage and/or heterogeneous data integration, there are still many unsolved problems and unexplored areas in computational pathology. In particular, digital pathology exhibits some unique challenges, such as high image dimension for whole-slide images, batch effects due to inconsistent imaging procedures, and often higher cost in data collection and annotation compared with other modalities of imaging data. These can worsen the situation where deep models are learned with limited data or integration of multi-modal data. By assembling a variety of viewpoints and a wealth of expertise, this Research Topic aims to offer a platform for computational pathology researchers to present cutting-edge methods and practical applications, and also for the field to promote collaborations among diverse groups.
This Research Topic invites AI experts, medical researchers, and physicians to contribute original research articles, reviews, and case studies, which focus on data/annotation-efficient modeling and integration of multi-modal data for computational pathology. The topics of interest include, but are not limited to:
- Weakly, active, semi- and self-supervised learning in digital pathology.
- Domain adaptation and generalization for cross-domain pathology data.
- Data synthesis including pathology image generation and translation.
- Federated learning to bridge pathology data silos.
- Multi-modal learning with different types of medical data.
- Geometric deep learning including graph neural networks for computational pathology.
Keywords:
Deep learning, Weakly-supervised learning, Self-supervised learning, Federated learning, Multi-modal learning, Generative models, Graph neural networks
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.