Introduction: Children with Autoimmune Liver Disease (AILD) may develop fibrosis-related complications necessitating liver transplant. We hypothesize that tissue-based analysis of liver fibrosis by Second Harmonic Generation (SHG) microscopy with Artificial Intelligence (AI) analysis can yield prognostic biomarkers in AILD.

Methods: Patients from single-center prospective studies with unstained slides from clinically-obtained liver biopsy at AILD diagnosis were identified. Baseline demographics and liver biochemistries were collected at diagnosis and 1 year. In collaboration with HistoIndex®, unstained slides underwent SHG/AI analysis to map fibrosis according to 9 quantitative fibrosis (qFibrosis) parameters.

Results: Sixty-four patients with AIH (50%), PSC (33%) or ASC (17%) at a median of 14 years old (range 3-24 years) were included. Four patients were listed for liver transplant within 36 months from diagnosis. An unsupervised analysis of qFibrosis parameters identified two distinct patient clusters: one with more females, PSC/ASC, and IBD. This group had more fibrosis by METAVIR classification (3 vs 2; p=0.0008) and persistently abnormal ALT at 12 months (42.4% vs 13.3%; p=0.01). qFibrosis parameters were most predictive of abnormal 12-month ALT. Perisinusoidal and overall fibrosis were most associated with abnormal 12-month biochemistries. In the multiple regression model, sex had an OR of 5.4 (p=0.02) and NumStr of 10.9 (p=0.0007) of abnormal 12-month ALT (AUC 0.82).

Conclusions: The application of SHG/AI algorithms in pediatric-onset AILD improves the predictive performance of liver histopathology. SHG allows to identify patients more at risk of severe liver disease as well as fibrosis patterns related to persistently abnormal biochemistries at one year from diagnosis.