Medical imaging is transforming thanks to the advent of foundation models, which promise to enhance adaptability and efficiency in data usage. Despite their potential, these models' performance varies when applied to differing clinical tasks and imaging modalities, such as histopathology, radiology, and ultrasound. One significant obstacle is the absence of universally accepted benchmarking standards, which results in inconsistent performance evaluations and complicates reproducibility.
This lack of standardised benchmarks hampers the acceptance and trust in these models' clinical applications, especially in crucial diagnostic procedures. A systematic and transparent evaluation is necessary to understand their capabilities, identify biases, and assess their generalizability. This is essential for advancing diagnostic synergies across various modalities.
This Research Topic aims to develop comprehensive, transparent benchmarking standards for foundation models in medical imaging. Across various modalities, including but not limited to histopathology, radiology, and ultrasound, the goal is to provide a framework that enhances model selection, optimisation, and adherence to regulatory requirements. The benchmarks will facilitate the comparison of model efficacy across tasks, including classification, segmentation, and prognosis, thereby contributing to more reliable and effective clinical implementations.
To gather further insights into benchmarking foundation models, we welcome articles addressing, but not limited to, the following themes:
• Development of benchmarking frameworks across various modalities, including histopathology, radiology, and ultrasound.
• Comprehensive cross-modal diagnostic tool evaluation.
• Prognostic accuracy assessments in various disease contexts.
• Comparative studies on biomarker detection efficiencies.
• Standards for tumour grading and progression analysis.
These themes aim to fortify the reliability of medical imaging analyses facilitated by foundation models and enhance their integration into clinical workflows.
Keywords: Histopathology, Medical Imaging, computational pathology, foundation models, pathomics, multiple instance learning, radiology, ultrasound
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.
