Artificial intelligence in digital pathology diagnosis and analysis: technologies, challenges, and future prospects

Abstract

Artificial intelligence (AI) offers transformative potential in pathology, where histopathological images remain the diagnostic gold standard due to their rich morphological and molecular information. While the rapid development of AI-driven computational pathology tools is revolutionizing disease interpretation, these technologies have not yet been systematically evaluated. Therefore, this review systematically evaluates AI applications across the diagnostic continuum, from image preprocessing and tumor classification to prognostic stratification and the discovery of predictive biomarkers. It presents a technical taxonomy of the algorithms and foundation models powering these applications, benchmarking their performance across diverse diagnostic tasks through rigorous comparative analyses. It also identifies critical challenges in clinical translation, including computational scaling, noisy annotations, interpretability gaps, and domain shifts. Finally, it proposes a roadmap for advancing AI applications in precision oncology and pathological research. By bridging technological innovation with clinical needs, this review aims to accelerate the integration of robust, unified, scalable AI solutions into diagnostic workflows.

Keywords: Artificial intelligence (AI); Pathology foundation model; Pathology images; Quantitative feature.

Fig. 1

Introduction to the digitized pathology image formation process. H&E hematoxylin and eosin, IHC immunohistochemistry, AI artificial intelligence

Fig. 2

AI applications in digital pathology primarily encompass basic pathology image processing, clinical screening and diagnosis, prognosis prediction, and biomarker discovery. The associated computational tasks and corresponding AI techniques are systematically summarized for each application. H&E hematoxylin and eosin, GANs generative adversarial networks, CNN convolutional neural network, VAE variational auto-encoders, ResNet residual network, U-Net U-shaped network, DNN deep neural network, GCN graph convolutional network

Fig. 3

Visualization of quantitative features at the whole-slide, tissue, and cell levels. The slide level includes percentage of cancerous areas (a), morphology of cancerous areas (b), TIL cluster density and structure (c), and tumor morphology and area, using computer-aided analysis to confirm the width, manual area, digital area, and morphological features of the primary tumor (d). The tissue level includes gland distribution and morphology (e), TME (f), tissue morphology and relationships (g), gland angles (h), local cellular interactions (i), and tissue abundance (j). The cell level includes nuclear structure and texture (k), intensity statistics and co-occurrence (l), nuclear morphology and topology (m), nuclear morphology and arrangement (n), quantitative nuclear features (o), nuclear shape and structure (p), using DL models to extract nuclear features followed by quantitative methods to extract features related to nuclear shape and structure; nuclear shape, texture, and orientation (q), nuclear cluster spatial map features (r), spatial distribution and topological features of cell clusters (s), and the size, shape, and texture features of cells (t). TIL tumor-infiltrating lymphocyte, TME tumor microenvironment

Fig. 4

Performance comparison of pathological large models across diagnosis classification tasks. a The bar chart illustrates the area under curves (AUCs) across multiple cancer datasets. b The bar chart illustrates the C-index on TCGA-LIHC and TCGA-LUAD datasets. Each model is represented by a different color. The results highlight the diagnostic capabilities of each model across diverse pathological contexts. HIPT hierarchical image pyramid transformer, UNI towards a general-purpose foundation model for computational pathology, CHIEF clinical histopathology imaging evaluation foundation, CONCH contrastive learning from captions for histopathology, TITAN transformer-based pathology image and text alignment network, TCGA The Cancer Genome Atlas, BRCA breast cancer, LUDA lung adenocarcinoma, RCC renal cell carcinoma, CRC colorectal cancer, MSI microsatellite instability, LIHC liver hepatocellular carcinoma