Development of AI Based Fibrosis Detection Algorithm by SHG/TPEF Microscopy for Fully Quantified Liver Fibrosis Assessment in MASH

Abstract

Background and aims: Metabolic dysfunction-associated steatotic liver disease (MASLD) is a major global cause of chronic liver disease, with the potential to progress from steatosis to metabolic dysfunction-associated steatohepatitis (MASH) and cirrhosis. Fibrosis is a key determinant of liver-related morbidity and mortality, highlighting the need for precise, reproducible assessment methods. This study aimed to develop and validate an Artificial Intelligence (AI)-based fibrosis detection algorithm using Second Harmonic Generation/Two Photon Excitation Fluorescence (SHG/TPEF) microscopy.

Methods: The algorithm integrates SHG/TPEF microscopy, which uses ultra-fast lasers to capture intrinsic optical signals from unstained liver biopsies, with Machine Learning (ML)-based image analysis. The resulting qFibrosis model quantifies collagen morphology to generate a continuous fibrosis index.

Results: A standardised workflow was established, encompassing sample acquisition, SHG/TPEF imaging, region-specific analysis and collagen feature quantification. Each step of the AI-based ML of qFibrosis algorithm used to assess and quantify liver fibrosis is described in detail in this study.

Conclusions: This AI-driven approach enables accurate, continuous quantification of liver fibrosis, overcoming the variability of traditional histopathology. The qFibrosis model has potential as a standardised tool for therapeutic evaluation and disease monitoring in MASLD/MASH, representing a significant advancement in liver fibrosis assessment.

Keywords: MASH; MASLD; Machine Learning; artificial intelligence; fibrosis.

FIGURE 1

Flowchart describing the process workflow.

FIGURE 2

Simplified Genesis200 imaging platform.

FIGURE 3

Examples of Laser Staining. (A) SHG/TPEF image for an unstained liver biopsy. The blue box is further expanded for detailed viewing in (B–D). (B) Shows TPEF microscopy‐based imaging (delineation of hepatocytes, indicated in red). (C) Shows SHG microscopy‐based identification of fibrosis (indicated in green). (D) Combined image.

FIGURE 4

Sequential execution of steps (A–J) included in ‘Machine Learning staining’. The details of each step are described in the subsequent passages.

FIGURE 5

(A) Illustrates all identified collagen fibres. (B) Intersections among branches are identified and quantified within the aggregated collagen fibres. (C) Connected component analysis is used to compute various collagen morphological features such as area, perimeter, length and width for aggregated and distributed fibres, respectively.