imageDx: NASH – A New Generation of Quantitative NASH Data

Traditional NAFLD Activity Scoring

Non-alcoholic fatty liver disease (NAFLD) is one of the most prevalent causes of liver dysfunction, with some estimates suggesting up to 30% of the general population as being affected (1). NAFLD, while mainly asymptomatic, can progress to non-alcoholic steatohepatitis (NASH) in a portion of individuals. NASH is characterized by multiple pathologic features: accumulation of fat within hepatocytes, infiltration of immune cells, presence of ballooning cells, and fibrosis. NASH is a progressive form of NAFLD with some individuals developing more serious chronic conditions such as cirrhosis and hepatocellular carcinoma, which may eventually necessitate liver transplantation. Given the increasing incidence of NASH in the general population and the potential progression to more serious liver conditions, it is crucial to develop methods to detect the early signs of NASH and definitively differentiate NAFLD and NASH diagnoses. imageDx: NASH is a new deep learning-based tool that allows researchers and pathologists to quantitatively assess NAFLD and fibrosis in human or rodent tissue. This data is highly accurate, reproducible and scalable for preclinical research, clinical trials and as a decision support tool for pathologists diagnosing patients.

NASH is a disease with a complicated pathophysiology. A NASH diagnosis depends on the presence, absence, and severity of several different tissue characteristics (2). The NAFLD Activity Score (NAS) is used to assess the NAFLD progression on a scale of 0-8 and is based on the severity of the following characteristics:

  • Steatosis (0-3): Excess lipid accumulation in cells, which displaces cell cytoplasm
  • Lobular Inflammation (0-3): Infiltration of immune cells in response to hepatocyte damage
  • Hepatocellular ballooning (0-2): Disruption of hepatocyte function and cell death
Examples of steatosis, lobular inflammation, and hepatocellular ballooning on images of H&E-stained liver tissue. The “mask” represents quantitative detection of these pathologic features by imageDx: NASH, Reveal’s deep learning-based NAS algorithm.

In addition to the NAFLD Activity Score, the progression of liver disease is assessed by determining the degree and location of fibrosis:

  • Fibrosis (0-4): Excess extracellular matrix deposition or scarring from aberrant tissue repair processes
Inter-pathologist variation between 9 pathologists reading 32 adult liver biopsy specimens (2).

A definitive NAFLD/NASH diagnosis is made by a pathologist visually assessing multiple pathologic features from a liver biopsy. The scoring system allows pathologists to combine information about various characteristics into a single, semi-quantitative value that provides insight into a patient’s condition. Inter-pathologist variability exists across all pathologic features when reading liver biopsies and results in inconsistent scoring, which may impact research or a patient’s diagnosis and proposed treatment plan (2). This variability, reflected in the low Kappa scores, underscores the need for a quantitative and reproducible scoring method for this disease. We have developed a quantitative approach to measuring the pathologies comprising the NAS score based on machine and deep learning that is precise and consistent, allowing accurate quantification across samples.

imageDx: NASH

imageDx: NASH is a deep learning-based tool that allows researchers and pathologists to quantitatively assess NAFLD and fibrosis in human or rodent tissue. These tools were co-developed with leading liver pathologists to ensure reliable data for research and clinical trials.

Left: Whole slide image of liver tissue stained with Masson’s Trichrome. Right: Quantitative fibrosis detection by imageDx: NASH.

imageDx: NASH accurately identifies and quantifies steatosis (micro- and macro-vesicular steatosis), lobular inflammation, and hepatocellular ballooning from whole slide images of H&E-stained liver tissue. The precise data generated from this approach is helpful to researchers working on rodent or human models of NASH or liver fibrosis, and a NAFLD Activity Score is also assigned based on these values. Similarly, a quantitative fibrosis score is generated from whole slide images of liver tissue stained by Masson’s Trichrome or Picrosirius Red. Quantitative data for the presence of Mallory bodies and iron deposition can also be generated.

Many early preclinical NAFLD studies are conducted using mouse/rat liver tissue. While good models in many aspects, rodent liver tissue differs qualitatively, quantitatively and physiologically from that of human in terms of lobules, hepatocellular ballooning, gene regulation and cell activation. Thus, we’ve developed separate NAFLD scoring algorithms for human and rodent liver to reflect those differences and generate accurate analysis for the respective tissue types.

Reveal offers a full-service histology laboratory to process, embed, section, stain, and image liver tissue samples, or our experienced team can analyze client-stained tissue slides or images. Quality control checkpoints at every step ensure trusted, actionable project results. Study deliverables include a comprehensive report, tabulated data, and access to whole slide images via the cloud-based imageDx platform. The imageDx platform is highly scalable, compatible with any image type, and fully compliant with medical standards for data security, making it the best way to view and analyze whole slide pathology images.

In summary, imageDx: NASH generates quantitative pathology data for steatosis (microvesicular and macrovesicular), inflammation, ballooning, and fibrosis. This data is highly accurate, reproducible and scalable for preclinical research, clinical trials, and as a decision support tool for pathologists diagnosing patients.

Contact us to learn how your study can benefit from AI-histopathology and quantitative image analysis.


  1. Dyson, Jessica K., Quentin M. Anstee, and Stuart McPherson. “Non-alcoholic fatty liver disease: a practical approach to diagnosis and staging.” Frontline gastroenterology 5.3 (2014): 211-218.
  2. Kleiner, David E., et al. “Design and validation of a histological scoring system for nonalcoholic fatty liver disease.” Hepatology 41.6 (2005): 1313-1321.


Jeremy Warner, Scientific Marketing Associate, Reveal Biosciences

Sara Becker-Catania, PhD, Associate Director, Product Development, Reveal Biosciences

Daniel Schroen, PhD, Commercial Lead, R&D, Reveal Biosciences

Pathology Revealed

Sign up to receive updates on the latest AI-powered pathology breakthroughs, access digital pathology resources, and more.