Artificial Intelligence (AI) has become popular in computer-aided decision-making in the healthcare domain. Many new AI methods are being developed for disease diagnosis and prognosis. Most diseases can be cured early and managed better if a timely diagnosis is done. The AI models can aid clinical diagnosis, thus they make the process efficient by reducing the workload of physicians and radiologists. However, the majority of AI methods rely on the use of single-modality data. For example, brain tumor detection uses brain MRI, skin lesion detection uses skin pathology images, and lung cancer detection uses lung CT or X-ray imaging. Single-modality AI models miss much-needed integration of complex features available from different modality data such as electronic health records (EHR), unstructured clinical notes, and different modalities of medical imaging – that otherwise form the backbone of clinical decision-making.
Many recent studies have demonstrated that the use of multimodal data tends to enhance the predictive performance of the AI models in medical imaging, e.g., through the leverage of diverse features in different modality data. Hence, multimodal AI models along with early/late data/inference fusion approaches can utilize complex features from data efficiently, thus resulting in better decisions. Most often, combining different modes of data is not straightforward as the data differ in dimensionality, modality, as well as availability. Exploring new methods to combine different data types often leads to new protocols and strategies for multimodal data collection, cleaning, pre-processing, and integration. Hence, many practical challenges exist in achieving the goal of training multimodal AI models for medical data for accurate prediction, explanation, etc.
Therefore, there is a big need to cover recent advancements and novel methods and techniques in multimodal AI for disease diagnosis, prognosis, or prevention. In this connection, this Research Topic will benefit both the AI and clinical researchers looking for new developments in multimodal AI methods in the medical data domain.
We welcome potential authors to submit original research articles, review articles, or research reports related to the scope of the Research Topic. The topics of interest include (but are not limited to):
• Multimodal AI for cancer diagnosis and prognosis including brain, lung, skin, colorectal, prostate, head and neck cancer.
• Multimodal AI for rare cancer
• Multimodal AI for neurodegenerative disorders.
• New methods and techniques to address data imbalance problems in multimodal AI.
• New methods and techniques for combining multimodality medical data for AI models.
• New methods for multimodality medical data analysis.
• Early fusion and late fusion approach for AI models in disease diagnosis.
• New methods of combining inferences and forecasts and applications in medical domains
• News methods and techniques for federated learning on multimodality medical data.
• Multimodal AI models training using medical images, omics, and EHR data for cancer diagnosis and prognosis.
• Techniques to overcome data imbalance challenges during the collection of different modality data.
• Quantization of neural networks for multimodal AI in medicine.
• Bayesian networks and probabilistic methods for effective medical decisions
• Multimodal AI for real-time medical image analysis using edge computing.
• Clinical transformation of explainable multimodal AI models.
• Surveys and reviews of recent studies on multimodal AI in disease diagnosis and prognosis.
• New software tools and medical decision systems of multimodal AI models
Artificial Intelligence (AI) has become popular in computer-aided decision-making in the healthcare domain. Many new AI methods are being developed for disease diagnosis and prognosis. Most diseases can be cured early and managed better if a timely diagnosis is done. The AI models can aid clinical diagnosis, thus they make the process efficient by reducing the workload of physicians and radiologists. However, the majority of AI methods rely on the use of single-modality data. For example, brain tumor detection uses brain MRI, skin lesion detection uses skin pathology images, and lung cancer detection uses lung CT or X-ray imaging. Single-modality AI models miss much-needed integration of complex features available from different modality data such as electronic health records (EHR), unstructured clinical notes, and different modalities of medical imaging – that otherwise form the backbone of clinical decision-making.
Many recent studies have demonstrated that the use of multimodal data tends to enhance the predictive performance of the AI models in medical imaging, e.g., through the leverage of diverse features in different modality data. Hence, multimodal AI models along with early/late data/inference fusion approaches can utilize complex features from data efficiently, thus resulting in better decisions. Most often, combining different modes of data is not straightforward as the data differ in dimensionality, modality, as well as availability. Exploring new methods to combine different data types often leads to new protocols and strategies for multimodal data collection, cleaning, pre-processing, and integration. Hence, many practical challenges exist in achieving the goal of training multimodal AI models for medical data for accurate prediction, explanation, etc.
Therefore, there is a big need to cover recent advancements and novel methods and techniques in multimodal AI for disease diagnosis, prognosis, or prevention. In this connection, this Research Topic will benefit both the AI and clinical researchers looking for new developments in multimodal AI methods in the medical data domain.
We welcome potential authors to submit original research articles, review articles, or research reports related to the scope of the Research Topic. The topics of interest include (but are not limited to):
• Multimodal AI for cancer diagnosis and prognosis including brain, lung, skin, colorectal, prostate, head and neck cancer.
• Multimodal AI for rare cancer
• Multimodal AI for neurodegenerative disorders.
• New methods and techniques to address data imbalance problems in multimodal AI.
• New methods and techniques for combining multimodality medical data for AI models.
• New methods for multimodality medical data analysis.
• Early fusion and late fusion approach for AI models in disease diagnosis.
• New methods of combining inferences and forecasts and applications in medical domains
• News methods and techniques for federated learning on multimodality medical data.
• Multimodal AI models training using medical images, omics, and EHR data for cancer diagnosis and prognosis.
• Techniques to overcome data imbalance challenges during the collection of different modality data.
• Quantization of neural networks for multimodal AI in medicine.
• Bayesian networks and probabilistic methods for effective medical decisions
• Multimodal AI for real-time medical image analysis using edge computing.
• Clinical transformation of explainable multimodal AI models.
• Surveys and reviews of recent studies on multimodal AI in disease diagnosis and prognosis.
• New software tools and medical decision systems of multimodal AI models