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

Comprehensive Summary

In this narrative review, the authors explore current applications of artificial intelligence tools across the digital pathology workflow, specifically for diagnostic, predictive, or analytical tasks based on whole-slide images. Through a discussion based on recent clinical and technical literature surveying machine learning and deep learning networks, the review examines tasks including preprocessing images for computer vision, tissue segmentation, tumor detection for identification and stratification, whole-slide imaging classification tasks for grading diagnosis, and risk assessment for stratifying patients into high-risk or low-risk groups based on several biomarkers. Instead of reporting original performance results for a machine learning–based clinical pathology platform, the authors argue that improved efficiency gains, reproducibility gains, and additional feature extraction capabilities for digital pathology have already been observed through AI-assisted analysis for clinicopathological diagnostics that have traditionally utilized human visual input for analysis.

Outcomes and Implications

When used in clinical environments, digital pathology enabled by AI has the potential to minimize variability in diagnosis, enhance efficiency in population pathology labs, and offer quantitative information not readily determined during manual histopathological examination. The proposed use of AI in tumor screening, grading help, risk modeling for prognosis, and quantification of biomarkers could aid in enhancing diagnosis consistency and enable pathologists to focus on particularly challenging cases. However, the authors do point out that at this stage, it is not advisable to apply current levels of AI for exclusive decision-making roles in clinical environments because of questions about generalization capabilities and readiness of AI under these aspects. When these issues come into focus, AI-augmented pathology would have an increasingly crucial supportive role in precision oncology for aiding risk and treatment planning while upholding pathology expert views as the gold clinical standard.