Badve Sunil, Kumar George L, Lang Tobias, Peigin Eli, Pratt James, Anders Robert, Chatterjee Deyali, Gonzalez Raul S, Graham Rondell P, Krasinskas Alyssa M, Liu Xiuli, Quaas Alexander, Saxena Romil, Setia Namrata, Tang Laura, Wang Hanlin L, Rüschoff Josef, Schildhaus Hans-Ulrich, Daifalla Khalid, Päpper Marc, Frey Patrick, Faber Felix, Karasarides Maria
Emory University School of Medicine, Atlanta, GA, USA.
Bristol Myers Squibb, Princeton, NJ, USA.
NPJ Precis Oncol. 2025 May 12;9(1):139. doi: 10.1038/s41698-025-00899-5.
Diagnostic certainty is the cornerstone of modern medicine and critical for maximal treatment benefit. When evaluating biomarker expression by immunohistochemistry (IHC), however, pathologists are hindered by complex scoring methodologies, unique positivity cut-offs and subjective staining interpretation. Artificial intelligence (AI) can potentially eliminate diagnostic uncertainty, especially when AI "trustworthiness" is proven by expert pathologists in the context of real-world clinical practice. Building on an IHC foundation model, we employed pathologists-in-the-loop finetuning to produce a programmed cell death ligand 1 (PD-L1) CPS AI Model. We devised a multi-head augmented reality microscope (ARM) system overlayed with the PD-L1 CPS AI Model to assess interobserver variability and gauge the pathologists' trust in AI model outputs. Using difficult to interpret regions on gastroesophageal biopsies, we show that AI-assistance improved case agreement between any 2 pathologists by 14% (agreement on 77% vs 91%) and among 11 pathologists by 26% (agreement on 43% vs 69%). At a clinical cutoff of PD-L1 CPS ≥ 5, the number of cases diagnosed as positive by all 11 pathologists increased by 31%. Our findings underscore the benefits of fully engaging pathologists as active participants in the development and deployment of IHC AI models and frame the roadmap for trustworthy AI as a bridge to increased adoption in routine pathology practice.
诊断确定性是现代医学的基石，对于实现最大治疗效益至关重要。然而，在通过免疫组织化学（IHC）评估生物标志物表达时，病理学家受到复杂评分方法、独特阳性临界值和主观染色解读的阻碍。人工智能（AI）有可能消除诊断不确定性，尤其是当AI的“可信度”在现实世界临床实践中得到专家病理学家证实时。基于一个IHC基础模型，我们采用了“病理学家参与”的微调方法来生成一种程序性细胞死亡配体1（PD-L1）CPS人工智能模型。我们设计了一个叠加有PD-L1 CPS人工智能模型的多头增强现实显微镜（ARM）系统，以评估观察者间的变异性，并衡量病理学家对人工智能模型输出结果的信任度。利用胃食管活检中难以解读的区域，我们发现人工智能辅助使任意两名病理学家之间的病例一致性提高了14%（一致性从77%提高到91%），在11名病理学家之间提高了26%（一致性从43%提高到69%）。在PD-L1 CPS≥5的临床临界值下，所有11名病理学家诊断为阳性的病例数量增加了31%。我们的研究结果强调了让病理学家作为积极参与者全面参与IHC人工智能模型的开发和部署的好处，并为可信人工智能制定了路线图，作为在常规病理学实践中提高采用率的桥梁。
