Bosch Jaime, Chung Chuhan, Carrasco-Zevallos Oscar M, Harrison Stephen A, Abdelmalek Manal F, Shiffman Mitchell L, Rockey Don C, Shanis Zahil, Juyal Dinkar, Pokkalla Harsha, Le Quang Huy, Resnick Murray, Montalto Michael, Beck Andrew H, Wapinski Ilan, Han Ling, Jia Catherine, Goodman Zachary, Afdhal Nezam, Myers Robert P, Sanyal Arun J
Department of Biomedical Research, University of Bern, Bern, Switzerland.
University of Barcelona-IDIBAPS and CIBERehd, Barcelona, Spain.
Hepatology. 2021 Dec;74(6):3146-3160. doi: 10.1002/hep.32087.
The hepatic venous pressure gradient (HVPG) is the standard for estimating portal pressure but requires expertise for interpretation. We hypothesized that HVPG could be extrapolated from liver histology using a machine learning (ML) algorithm.
Patients with NASH with compensated cirrhosis from a phase 2b trial were included. HVPG and biopsies from baseline and weeks 48 and 96 were reviewed centrally, and biopsies evaluated with a convolutional neural network (PathAI, Boston, MA). Using trichrome-stained biopsies in the training set (n = 130), an ML model was developed to recognize fibrosis patterns associated with HVPG, and the resultant ML HVPG score was validated in a held-out test set (n = 88). Associations between the ML HVPG score with measured HVPG and liver-related events, and performance of the ML HVPG score for clinically significant portal hypertension (CSPH) (HVPG ≥ 10 mm Hg), were determined. The ML-HVPG score was more strongly correlated with HVPG than hepatic collagen by morphometry (ρ = 0.47 vs. ρ = 0.28; P < 0.001). The ML HVPG score differentiated patients with normal (0-5 mm Hg) and elevated (5.5-9.5 mm Hg) HVPG and CSPH (median: 1.51 vs. 1.93 vs. 2.60; all P < 0.05). The areas under receiver operating characteristic curve (AUROCs) (95% CI) of the ML-HVPG score for CSPH were 0.85 (0.80, 0.90) and 0.76 (0.68, 0.85) in the training and test sets, respectively. Discrimination of the ML-HVPG score for CSPH improved with the addition of a ML parameter for nodularity, Enhanced Liver Fibrosis, platelets, aspartate aminotransferase (AST), and bilirubin (AUROC in test set: 0.85; 95% CI: 0.78, 0.92). Although baseline ML-HVPG score was not prognostic, changes were predictive of clinical events (HR: 2.13; 95% CI: 1.26, 3.59) and associated with hemodynamic response and fibrosis improvement.
An ML model based on trichrome-stained liver biopsy slides can predict CSPH in patients with NASH with cirrhosis.
肝静脉压力梯度（HVPG）是评估门静脉压力的标准，但解读需要专业知识。我们假设可以使用机器学习（ML）算法从肝脏组织学推断HVPG。
纳入了来自一项2b期试验的非酒精性脂肪性肝炎（NASH）合并代偿期肝硬化患者。对基线、第48周和第96周的HVPG和活检样本进行集中审查，并使用卷积神经网络（PathAI，马萨诸塞州波士顿）对活检样本进行评估。在训练集（n = 130）中使用三色染色的活检样本，开发了一个ML模型以识别与HVPG相关的纤维化模式，并在一个保留测试集（n = 88）中对所得的ML HVPG评分进行验证。确定ML HVPG评分与测量的HVPG和肝脏相关事件之间的关联，以及ML HVPG评分对临床显著门静脉高压（CSPH）（HVPG≥10 mmHg）的表现。ML-HVPG评分与HVPG的相关性比形态计量学测定的肝胶原更强（ρ = 0.47对ρ = 0.28；P < 0.001）。ML HVPG评分能够区分HVPG正常（0 - 5 mmHg）、升高（5.5 - 9.5 mmHg）和CSPH的患者（中位数：1.51对1.93对2.60；所有P < 0.05）。在训练集和测试集中，ML-HVPG评分对CSPH的受试者工作特征曲线下面积（AUROCs）（95% CI）分别为0.85（0.80, 0.90）和0.76（0.68, 0.85）。通过添加结节性、增强肝纤维化、血小板、天冬氨酸转氨酶（AST）和胆红素的ML参数，ML-HVPG评分对CSPH的辨别能力得到改善（测试集中的AUROC：0.85；95% CI：0.78, 0.92）。尽管基线ML-HVPG评分无预后价值，但变化可预测临床事件（HR：2.13；95% CI：1.26, 3.59），并与血流动力学反应和纤维化改善相关。
基于三色染色肝活检切片的ML模型可预测NASH合并肝硬化患者的CSPH。
