Boston, Massachusetts – PathAI, a global provider of AI-powered technology applied to pathology research, today announced the results of multiple applications of their platform to the quantitative assessment of liver disease. PathAI machine learning (ML) models were trained to accurately predict histological features of liver disease severity including fibrosis, steatosis, ballooning, and inflammation across 3 liver diseases: NASH, chronic HBV, and PSC. In all cases, models were trained on digitized, whole slide images of hematoxylin and eosin (H&E), trichrome, and/or picrosirius red-stained liver biopsies.
“Accurate and reproducible assessment of liver histology using AI-powered pathology has the potential to generate new insights into the pathologic mechanisms underlying a diverse set of liver diseases and to provide a basis for the quantitative assessment of the effect of treatment on disease progression,” said PathAI co-founder and Chief Executive Officer Andy Beck MD, PhD. “It is exciting to work with long-standing partners, like Gilead, to demonstrate the potential of the PathAI platform, especially for supporting research and development efforts to bring new therapies to patients.”
The new results are presented in three poster sessions and an oral presentation at The Digital International Liver Congress™ which will be held virtually from 27-29 August.
ML-based analysis of data from a Phase 2b trial of combination therapies for the treatment of advanced fibrosis (F3-F4) due to NASH (ATLAS) demonstrated regression of fibrosis and improvement in other histologic features of NASH in patients treated with the combination of cilofexor and firsocostat compared with placebo. - Patients from the STELLAR trials with progression to cirrhosis or liver-related clinical events were shown to have greater areas of portal inflammation and hepatocellular ballooning, and higher fibrosis scores at baseline (BL) based on quantitative, ML-based assessment. HBV-infected patients with chronic HBV infection that did not have cirrhosis regression following long-term tenofovir disoproxil fumarate (TDF) therapy were shown by ML-based pathology to have histologic features consistent with underlying NASH. ML models in patients with PSC identified features at BL, including fibrosis scores and interface inflammation, that correlated with risk of PSC-related clinical events. “The data generated with PathAI demonstrate the potential of their automated, ML-based approach to evaluation of histology in clinical trials of therapies for liver diseases,” said Rob Myers, MD, Vice President and Liver Fibrosis Clinical Research Lead, Gilead Sciences. “We look forward to continuing our partnership with PathAI to explore the application of their deep learning platform to identify novel histologic features associated with treatment response and disease progression in additional trials.” The Phase 2b ATLAS trial investigated combination therapies, including cilofexor, firsocostat, and selonsertib, for treatment of patients with advanced fibrosis (F3-F4) due to NASH. While manual histologic assessment did not identify statistically significant increases in ≥1-stage fibrosis improvement without worsening of NASH, ML-based quantification of fibrosis demonstrated a reduction in ML NASH Clinical Research Network (CRN) fibrosis score and a shift to less advanced fibrosis patterns in patients treated with the combination of cilofexor and firsocostat versus placebo. Oral Presentation LBO04 Safety and efficacy of combination therapies including cilofexor/firsocostat in patients with bridging fibrosis and cirrhosis due to NASH: Results of the Phase 2b ATLAS trial In 1,593 NASH patients with F3 or F4 fibrosis participating in two Phase 3 trials evaluating selonsertib, ML models identified features at BL that were predictive of progression to cirrhosis or a liver-related clinical event. Disease progression was associated with higher ML NASH CRN and Ishak fibrosis scores, as well as greater proportionate areas of hepatocellular ballooning and portal inflammation. Among patients with cirrhosis (F4), liver-related clinical events were also associated with a higher proportionate area of bile ducts and lower proportionate area of steatosis. Poster FRI003 Machine Learning Models Identify Novel Histologic Features Predictive of Clinical Disease Progression in Patients With Advanced Fibrosis Due to Nonalcoholic Steatohepatitis In 330 patients with chronic HBV infection participating in two registrational trials of TDF, ML-Ishak scoring illustrated the heterogeneity of fibrosis within tissue, a feature that is not captured with manual histologic staging. ML-Ishak scoring also revealed fibrosis regression at year 1 that was only detectable at year 5 with manual assessment. Significantly, BL ML steatosis scores were higher in patients that did not have cirrhosis regression at year 5 compared with those with cirrhosis regression, suggesting underlying NASH in the former patients. Poster LBP31 Machine Learning Identifies Histologic Features Associated With Regression of Cirrhosis in Treatment for Chronic Hepatitis B Among 141 patients with PSC participating in a clinical trial of simtuzumab, ML-Ishak scoring strongly correlated with staging of fibrosis by an expert central pathologist. The ML models were prognostic, showing that higher ML Ishak scores or a greater area of inflammation at BL were associated with an increased risk of PSC-related clinical events during follow-up. Poster FRI173 Machine Learning Models Accurately Interpret Liver Histology and Are Associated With Disease Progression in Patients With Primary Sclerosing Cholangitis
PathAI is a leading provider of AI-powered research tools and services for pathology. PathAI’s platform promises substantial improvements to the accuracy of diagnosis and the efficacy of treatment of diseases like cancer, leveraging modern approaches in machine and deep learning. Based in Boston, PathAI works with leading life sciences companies and researchers to advance precision medicine. To learn more, visit pathai.com.