Translational AI in whole-slide image cancer histopathology: state of the art and regulatory-approved solutions

With an emphasis on applied evidence, we present an in-depth evaluation of artificial intelligence (AI) in cancer histopathology through the lens of United States Food and Drug Administration (FDA)-approved and European Conformity-marked whole-slide image In Vitro Diagnostic Medical Devices. Having identified only four existing FDA-approved whole-slide image cancer solutions for a narrow range of applications, we conclude that AI in digital histopathology remains in an emerging state. Best practices were identified by examining development and validation evidence across market-approved solutions. Findings were contrasted with state-of-the-art research-only AI histopathology pipelines. Insights were drawn regarding applications, learning modalities, processing strategies, statistical methods, and validation approaches. Regulatory guidelines were evaluated from FDA and UK Government documentation as well as academic literature, with patient safety highlighted as a central concern. Approved products were observed to integrate efficiently into existing clinical decision-making frameworks, with future potential to enhance the use of pathologist consensus in AI applications. Biomarker assays may be coupled specifically to emerging therapies, but challenges remain for direct clinical adoption outside the research-only sphere. Although hurdles remain in validating agentic and generative AI for medicine, further adoption of state-of-the-art algorithmic frameworks-including transformer architectures and multimodal approaches-is anticipated. As pan-cancer systems emerge, computational modules and engineering principles may also be retargeted to underrepresented use cases. Consequently, this review provides a forward-looking framework for translational, market-relevant histopathology AI.