Neuschwander-Tetri Brent A, Akbary Kutbuddin, Carpenter Danielle H, Noureddin Mazen, Alkhouri Naim
Division of Gastroenterology and Hepatology, Saint Louis University, St. Louis, MO, 63110, USA.
HistoIndex, Teletech Park, 20 Science Park Road, Singapore 117674, Singapore.
J Hepatol. 2025 Apr 30. doi: 10.1016/j.jhep.2025.04.026.
Conventional histopathological evaluation of liver biopsy slides has been invaluable in assessing the causes of liver injury, the severity of the underlying disease processes, and the degree of resulting fibrosis. However, the use of conventional histologic assessments as endpoints in clinical trials is limited by the reliability of scoring systems, variability in interpretation of histologic features and translation of continuous variables into categorical scores. To increase the precision and reproducibility of liver biopsy assessment, several artificial intelligence/machine learning (AI/ML) approaches have been developed to analyse high-resolution digital images of liver biopsy specimens. Multiple AI/ML platforms are in development and promising results have been reported in post hoc analyses of clinical trial biopsies. One such technique employs second harmonic generation/two photon excitation (SHG/TPE) microscopy, which is unique in using unstained liver biopsies (avoiding challenges related to staining variability), to provide high-resolution images of collagen fibres, enabling assessment and quantification of collagen morphometry. One SHG/TPE microscopy methodology coupled with AI/ML-based analysis, qFibrosis, has been used post hoc as an exploratory endpoint in several clinical trials for MASH (metabolic dysfunction-associated steatohepatitis), which have demonstrated its ability to provide a consistent and more nuanced assessment of liver fibrosis that still correlates well with traditional staging. In this review, we summarise the development of qFibrosis and outline the need for additional studies to validate it as a sensitive marker of changes in fibrosis in the context of treatment trials and to correlate these changes with subsequent liver-related outcomes.
肝活检切片的传统组织病理学评估在评估肝损伤原因、潜在疾病进程的严重程度以及由此产生的纤维化程度方面一直具有重要价值。然而，在临床试验中使用传统组织学评估作为终点受到评分系统可靠性、组织学特征解释的变异性以及连续变量转化为分类评分的限制。为了提高肝活检评估的准确性和可重复性，已经开发了几种人工智能/机器学习（AI/ML）方法来分析肝活检标本的高分辨率数字图像。多个AI/ML平台正在开发中，并且在临床试验活检的事后分析中报告了有前景的结果。一种这样的技术采用二次谐波产生/双光子激发（SHG/TPE）显微镜，它在使用未染色的肝活检标本方面独具特色（避免了与染色变异性相关的挑战），以提供胶原纤维的高分辨率图像，从而能够评估和量化胶原形态学。一种结合了基于AI/ML分析的SHG/TPE显微镜方法，即qFibrosis，已在事后被用作多项关于代谢功能障碍相关脂肪性肝炎（MASH）的临床试验中的探索性终点，这些试验证明了它能够对肝纤维化提供一致且更细致入微的评估，并且仍然与传统分期有很好的相关性。在这篇综述中，我们总结了qFibrosis的发展，并概述了开展更多研究的必要性，以验证它作为治疗试验背景下纤维化变化的敏感标志物，并将这些变化与随后的肝脏相关结局相关联。
