From histopathology images to molecular characterisation of tumours: The artificial intelligence path.
Year: 2024
Type: book-chapter
Abstract: Artificial neural networks regained the spotlight of AI research efforts and are now themain driving force behind spectacular developments in a wide range of domains. Formedical imaging in general and pathological imaging in particular, these advancementsopen a whole new universe of applications. Molecular data represents a natural companionto imaging and the results briefly reviewed in this chapter demonstrate some of thepossibilities of such explorations.Currently, apart from MSI endpoint, the performance of the models predictingmolecular features, while promising, does not warrant their application in clinicalscenarios. This does not only necessarily indicate a limitation of the network architecturesemployed, but also issues with the data used in training the models.Transitioning the AI models for pathology from research laboratory to clinical practicestill requires extensive validation and faces several major challenges from more technical(performance and stability across platforms) to interpretability and regulatory. But,the question is not whether these AI companions will become ubiquitous, but rather‘when?’.
Source: HAL
Authors: Popovici, Vlad, Racoceanu, Daniel
Language: English
Topic: AI in cancer detection
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W4406586033
