人工智能数字病理学为 NASH 纤维化回归提供了新见解

StephenW

人工智能数字病理学为 NASH 纤维化回归提供了新见解
斯蒂芬·帕迪拉
23 小时前
AddThis 共享按钮
分享到谷歌书签
分享到 FacebookShare to TwitterShare to PrintShare to More
人工智能数字病理学为 NASH 纤维化回归提供了新见解

具有人工智能 (AI) 分析的二次谐波产生/双光子激发荧光 (SHG/TPEF) 显微术揭示了非酒精性脂肪性肝炎 (NASH) 中治疗诱导的纤维化消退的新方面，现有分期系统通常无法捕获这些方面，报告称 学习。

“带有 AI 分析的 SHG/TPEF 显微术提供了对 NASH 特征的标准化评估，尤其是连续规模的肝纤维化和胶原纤维定量，”研究人员说。 “这种方法被用于深入了解在参与 FLIGHT-FXR 研究的患者中使用非胆汁酸法尼醇 X 受体激动剂 tropifexor (TXR) 治疗后的纤维化动力学。”

接受安慰剂 (n=34)、TXR 140 μg (n=37) 或 TXR 200 μg (n=28) 治疗 48 周的 99 名 NASH（纤维化分期 F2 或 F3）患者共进行了 198 次肝活检 （配对：基线和治疗结束）。 然后检查来自这些活组织检查的未染色切片。

研究人员使用 SHG/TPEF 显微镜来量化肝纤维化、肝脂肪和膨胀的肝细胞。 他们还定量评估了肝小叶内隔膜形态、胶原纤维参数和区域分布的变化。

与传统显微镜不同，数字分析显示基线时 F2 或 F3 纤维化患者的整体肝纤维化与治疗相关减少，窦周纤维化显着消退。 [J Hepatol 2022;77:1399-1409]

在同时对纤维化和脂肪变性进行区域定量时，肝脂肪减少较多的患者也显示窦周纤维化减少最多。 最后，研究人员注意到隔膜形态的倒退变化和隔膜参数的下降几乎完全发生在 F3 患者身上，这些患者在常规评分中被认为“未改变”。

总的来说，这项研究表明，使用数字病理学和 SHG/TPEF 衍生结果在 NASH 中提供了两个新特征。

“[F] 首先，它提供了对治疗诱导的纤维化消退的新理解，首先是肝窦周围纤维化显着减少，以响应肝细胞中脂肪和脂毒性驱动因子的减少，随后延伸到门静脉纤维化，”研究人员说。

“[S]其次，它通过揭示 NASH 临床研究网络评分系统和传统显微镜未捕获的 TXR 的抗纤维化作用，证明了人工智能数字病理学的优势，”他们补充说。

数字方法

以前的研究还使用 SHG/TPEF 显微镜检查乙型肝炎和非酒精性脂肪肝的纤维化变化。 [J Hepatol 2014；61：260-269； 肝病学 2020；71：1953-1966； 科学报告 2018；8：2989； 肠道 2020；69：1116-1126； Clin Mol Hepatol 2021;27:44-57]

除了 SHG/TPEF 显微镜外，还开发了其他数字方法，需要染色幻灯片。 这些方法使用监督或半监督机器学习模型评估预定义的 NASH 特征。 [Clin Mol Hepatol 2021；27：44-57； Ann Diagn Pathol 2020;47151518; 实验室投资 2020；100:147-160； 新陈代谢 2021；117154707]

研究人员说：“总的来说，这些研究强调了 AI 数字病理学为肝病尤其是 NASH 研究带来的创新和品质。”

“事实上，人工智能数字病理学有望被整合到诊断和研究组织病理学的工作流程中，以符合人工智能在医学中越来越多使用的总体趋势，”他们补充道。 [肝病学 2020；72：2000-2013； Clin Res Hepatol Gastroenterol 2020；44：1-3； J Clin Pathol 2021;74:448-455; J Hepatol 2019;70:1016-1018]

“肝纤维化是 NASH 临床结果的关键预后决定因素。 目前的评分系统存在局限性，尤其是在评估纤维化退化方面，”研究人员指出。

StephenW

Digital pathology with AI offers new insights into fibrosis regression in NASH
Stephen Padilla
23 hours ago
AddThis Sharing Buttons
Share to Google Bookmark
Share to FacebookShare to TwitterShare to PrintShare to More
Digital pathology with AI offers new insights into fibrosis regression in NASH

Second harmonic generation/two-photon excitation fluorescence (SHG/TPEF) microscopy with artificial intelligence (AI) analyses reveals novel facets of treatment-induced fibrosis regression in nonalcoholic steatohepatitis (NASH), which are not normally captured by existing staging systems, reports a study.

“SHG/TPEF microscopy with AI analyses provides standardized evaluation of NASH features, especially liver fibrosis and collagen fiber quantitation on a continuous scale,” the researchers said. “This approach was applied to gain in-depth understanding of fibrosis dynamics after treatment with tropifexor (TXR), a nonbile acid farnesoid X receptor agonist in patients participating in the FLIGHT-FXR study.”

Ninety-nine patients with NASH (fibrosis stage F2 or F3) who received placebo (n=34), TXR 140 μg (n=37), or TXR 200 μg (n=28) for 48 weeks underwent a total of 198 liver biopsies (paired: baseline and end-of-treatment). Unstained sections from these biopsies were then examined.

The researchers used SHG/TPEF microscopy to quantify liver fibrosis, hepatic fat, and ballooned hepatocytes. They also quantitatively assessed changes in septa morphology, collagen fibre parameters, and zonal distribution within liver lobules.

Unlike conventional microscopy, digital analyses revealed treatment-related reductions in overall liver fibrosis and noticeable regression in perisinusoidal fibrosis in patients with F2 or F3 fibrosis at baseline. [J Hepatol 2022;77:1399-1409]

On concomitant zonal quantitation of fibrosis and steatosis, patients with greater hepatic fat reduction also showed the greatest reduction in perisinusoidal fibrosis. Finally, the researchers noted regressive changes in septa morphology and decrease in septa parameters almost exclusively in F3 patients, who were deemed “unchanged” with conventional scoring.

Overall, this study showed that using digital pathology with SHG/TPEF-derived results provided two novel features in NASH.

“[F]irstly, it provides a new understanding of treatment-induced fibrosis regression starting with marked reduction in perisinusoidal fibrosis in response to decreased fat and lipotoxic drivers in hepatocytes that subsequently extends to portal fibrosis,” the researchers said.

“[S]econdly, it demonstrates the advantages of AI digital pathology by revealing antifibrotic effects of TXR which were not captured by the NASH clinical research network scoring system and conventional microscopy,” they added.

Digital methods

Previous studies have also used SHG/TPEF microscopy to examine fibrosis changes in hepatitis B and nonalcoholic fatty liver disease. [J Hepatol 2014;61:260-269; Hepatology 2020;71:1953-1966; Sci Rep 2018;8:2989; Gut 2020;69:1116-1126; Clin Mol Hepatol 2021;27:44-57]

Apart from SHG/TPEF microscopy, other digital methodologies have been developed, requiring stained slides. These approaches assessed predefined NASH features with supervised or semi-supervised machine-learning models. [Clin Mol Hepatol 2021;27:44-57; Ann Diagn Pathol 2020;47151518; Lab Invest 2020;100:147-160; Metabolism 2021;117154707]

“Collectively, these studies emphasize the innovation and qualities that AI digital pathology brings to investigations of liver diseases, particularly NASH,” the researchers said.

“Indeed, AI digital pathology is expected to be integrated into the workflow of diagnostic and research histopathology in line with the general trend toward the increasing use of AI in medicine,” they added. [Hepatology 2020;72:2000-2013; Clin Res Hepatol Gastroenterol 2020;44:1-3; J Clin Pathol 2021;74:448-455; J Hepatol 2019;70:1016-1018]

“Liver fibrosis is a key prognostic determinant for clinical outcomes in NASH. Current scoring systems have limitations, especially in assessing fibrosis regression,” the researchers noted.