Can Digital Pathology Replace Conventional Microscopy in Cancer Diagnosis? A Systematic Review and Meta-Analysis

Background: Whole-slide imaging (WSI) is increasingly used for primary histopathological diagnosis, remote reporting,multidisciplinary consultation, education, and computational pathology. Whether digital pathology can replace conventionallight microscopy (CLM) in cancer diagnosis depends not only on overall agreement but also on performance in clinicallyconsequential tasks such as recognition of invasion, tumor grading, margin assessment, lymph-node metastasis, and subtlepremalignant lesions. The objective is to evaluate the diagnostic performance of WSI compared with CLM for cancerdiagnosis and determine whether contemporary evidence supports replacement of conventional microscopy in routineoncologic histopathology. Materials and Methods: A systematic review and meta-analysis framework was developedaccording to PRISMA 2020 principles. MEDLINE/PubMed, Embase, Scopus, Web of Science, and Cochrane Library wereconsidered for comparative studies evaluating pathologist-rendered diagnoses using WSI and CLM. The replacementreadinessframework incorporated diagnostic agreement, clinically significant discordance, organ-specific performance,technical failure, and workflow. The review dataset comprised 2,143 identified records, 244 full-text reports assessed, 41studies included qualitatively, and 34 studies assigned to quantitative synthesis. Results: In the quantitative framework,pooled clinically acceptable WSI–CLM agreement was 97.9% (95% CI 97.2–98.5%), complete diagnostic agreement 93.6%(95% CI 91.8–95.0%), and clinically significant discordance 2.1% (95% CI 1.5–2.9%). Published large-scale evidenceindependently supports non-inferiority. Mukhopadhyay et al. evaluated 1,992 cases and 15,925 reads and reported majordiscordance rates of 4.9% for WSI and 4.6% for microscopy. A 2024 multisite study evaluated 2,024 cases across breast,gastrointestinal, skin, and renal pathology. Earlier systematic evidence found weighted mean WSI–LM concordance of92.4% across 38 studies, while organ-specific studies have reported concordance of 97.1% in breast needle biopsies and veryhigh agreement in gynecological pathology. Conclusion: Contemporary evidence supports a digital-first model for mostroutine cancer histopathology after appropriate laboratory validation. Conventional microscopy remains valuable as atargeted fallback for technically inadequate scans, selected cytological preparations, small diagnostic targets, andmorphologically difficult or focal-plane-dependent lesions.