African Journal of Biomedical Sciences
Editor-in-Chief: Dr. Abdelmonem Awad Hegazy | ISSN: 3136-8026 | Frequency: Biannual | Publication Format: Open Access | Language: English | Indexing/Listing :

Past Issues of African Journal of Biological Sciences

Volume 2, Issue 1, January 2026
Review Article

From whole-slide images to multimodal diagnostics: Foundation models, spatial omics and generative AI in pathology and laboratory medicine

| Open Access

Avery Williams1*
Afj.Bio.Med.Sci. 2(1) (2026) 17-26,https://doi.org/10.62587/AFJBMS.2.1.2026.17-26
Received: 19/09/2025|Accepted: 30/11/2025|Published: 25/01/2026

Abstract

Digital pathology, alongside the revolution of Artificial Intelligence (AI), molecular profiling, spatial omics, liquid biopsy, and multimodal computational systems, is revolutionizing both pathology and laboratory medicine. Conventional pathology is based mainly on the generation of morphological information, and only the interpretation of this morphological information is the expertise of a specialist, while in laboratory medicine it is possible to generate large amounts of biochemical, haematological, microbiological and molecular information. By combining these data streams one can make the shift from single instances of diagnosis towards a more complete, computer aided disease characterisation. The acquisition of WSIs has paved the way for machine-learning solutions for tissue segmentation, tumor detection, grading, biomarker quantification and prediction of prognosis. But, traditional image category-specific models are still constrained by the need for annotations, lack of cross-scan or cross-staining transfer, and also domain variations, which come from different scanner types, staining, tissueprep, and patient types. Pathology Foundation Models like UNI, CONCH, Virchow, Prov-GigaPath provide transferable representations that can aid various downstream tasks. Generative and multimodal AI also facilitates the integration of histological images with the pathology report, clinical information and molecular findings. Spatial transcriptomics and proteomics enable providing molecular information in the anatomic context, and liquid biopsy can be used to allow for minimal invasive longitudinal monitoring. However, there are challenges to clinical translation, such as bias, interpretability, automatism bias, cyber security, regulation and replicability. In this narrative review, the converging technologies and the opportunities for re-shaping precision diagnostics are discussed. This new paradigm is focused
on AI-enabled rather than AI-depleted patho-lab skills 


Keywords: Digital pathology, Computational pathology, Artificial intelligence, Foundation models, Generative AI, Spatial omics, Laboratory medicine, Liquid biopsy, Precision diagnostics, Multimodal AI

 Full text  Download  XML