Trustworthy AI for Medical Imaging: Advanced Modeling and Clinical Prediction

Medical imaging is a process of visualizing the tissues, organs and structure of the human body to diagnose, monitor, and treat medical conditions. It plays a vital role in precision medicine. In current practice, disease diagnosis using medical imaging is mainly based on visual information in the images and relies on the experience of physicians who interpret the visual information. But visual information and physician experience include limitations and uncertainties, which affect the accuracy and efficiency of medical imaging. Recently, artificial intelligence (AI) has been introduced to improve the accuracy and efficiency of medical imaging by using advanced algorithms and models such as deep learning models. As a matter of fact, AI is transforming medical imaging by leveraging large-scale data to support tasks such as image reconstruction, segmentation, registration, detection, and reporting. AI is also able to discover unvisual information or features contained in medical images utilizing omics. Such information can be used not only for diagnosis and treatment but for quantitative prediction of clinical outcomes. While AI application in medicine has been shown promising, it is in the early stage and has not significantly impacted clinical practice. To achieve meaningful clinical impact, AI models must be not only accurate and efficient, but also trustworthy, robust, and safe in real-world practice.

This Research Topic focuses on trustworthy and efficient AI modeling for medical imaging analysis and clinical outcome prediction. We are especially interested in studies that connect algorithmic innovation to clinical relevance, workflow efficiency, and safety. This Research Topic aims to move beyond isolated algorithmic improvements by emphasizing integration of AI methods to provide coherence within healthcare settings. By highlighting both foundational developments and downstream applications, this Research Topic seeks to offer current cutting edge AI studies for researchers, clinicians, and industry partners working at the interface of AI and medical imaging.

We welcome both original research articles and review papers. Research areas may include, but are not limited to, the following:

- AI for image reconstruction, enhancement, segmentation, registration, detection, and decision support
- Multimodal models for medical imaging that incorporate complementary clinical data (e.g., EHRs, omics, and biological data)
- Large language models, vision language models, and foundation models for medical imaging interpretation, reporting, and workflow support
- Generative models for image synthesis, data augmentation, and cross-modality translation
- Model robustness, uncertainty estimation, and out-of-distribution detection
- Explainable and interpretable AI, including bias assessment, fairness, and equity in imaging AI
- Quality assurance and quality control frameworks, and safety science for AI in medical imaging
- Comparisons of prediction models and algorithms
- Predictors of clinical outcomes for different imaging modalities
- Accuracies of outcome predictions for various diseases