Disclaimer
*The AI-based algorithm for automated PEC developed by Alimentiv is not a medical device and is not used by healthcare professionals for clinical diagnosis or for the provision of patient care. This tool is intended for research use only (RUO).

Alimentiv is developing digital image analysis tools to aid pathologists in scoring patient biopsies and improve the histopathological assessment of eosinophilic esophagitis (EoE) samples in clinical trials. For patients in clinical trials with EoE, timely histological assessment is the first step towards understanding and managing the chronic symptoms they experience, such as dysphagia, food impaction, chest pain, and regurgitation.1

We are actively shaping the design of clinical trials to support the development of new therapies for eosinophil-associated gastrointestinal disorders (EGID), a group of conditions that includes EoE. Despite its increasing prevalence, there are currently limited approved therapies in the US for treating EGID. Histologic assessment, such as peak eosinophil count (PEC), is currently the gold standard for EoE and EGID diagnosis. Alimentiv is committed to improving clinical trials with innovative tools supporting histologic scoring systems to better understand and characterize both EoE and other EGIDs, including gastritis (EoG), enteritis (EoN), and colitis (EoC).

What is eosinophilic esophagitis?

Eosinophilic esophagitis is a chronic, immune-mediated condition of the esophagus characterized by mucosal eosinophil infiltration. This infiltration triggers inflammation, structural damage, and dysfunction of the esophagus.1,2 Eosinophils are common, specialized immune cells with varied functions. They are implicated in many inflammatory processes, including modulating responses, particularly those related to allergic reactions.

In the case of EoE, the excessive presence of these cells in the esophagus is thought to cause unresolved inflammation and remodeling of the esophagus, leading to fibrosis, strictures, stiffening, and dysmotility. The severity and outcome of EoE can be proportional to the degree of mucosal infiltration by eosinophils, affirming the critical role these cells play in the disease.3 In clinical practice, diagnosis is made based on a combination of patient-reported symptoms, endoscopic assessment, and histologic evaluation of esophageal biopsies, with the later considered the gold standard approach for EoE disease severity grading.

Histological assessment challenges and Alimentiv’s solution

The main criterion for histologic evaluation of EoE is a peak eosinophil count (PEC) of 15 or more intraepithelial eosinophils within at least one high power field (HPF).4,5 Eosinophils are visually unique when stained by H&E and are identifiable by their distinctive bilobed nucleus and characteristic pink cytoplasmic granules. The presence of 15 or more eosinophils per HPF is a highly accurate indicator of EoE, boasting a sensitivity of 100% and specificity of 96%.

During EoE clinical trials, pathologists are required to score esophageal biopsies for PEC which involves analyzing six to nine biopsies from the proximal and distal esophagus, resulting in a significant number of histological slides for pathologists to manually evaluate. This workload becomes even more substantial in longitudinal clinical studies. The manual counting of hundreds of eosinophils per slide can be both time-consuming and labor-intensive.

To address these challenges, Alimentiv’s precision medicine team has developed an AI-based reading tool to assist pathologists in automatically quantifying the peak eosinophil count.6

AI-powered image analysis for improved EoE histology

Alimentiv recently presented its new AI-based algorithm for automated PEC quantification in EoE biopsies at Digestive Disease Week.7

The team developed a deep-learning algorithm by using a set of H&E-stained esophageal biopsy images from subjects with a wide range of EoE disease severity. A set of test images and associated eosinophil manual annotations were used to train the algorithm to automatically quantify intraepithelial eosinophils across the entire biopsy section, generate heatmaps that highlight the densest eosinophil areas (i.e., HPFs), and estimate PEC in those areas (Figures 1 & 2).

Figure 1. A) Image of hematoxylin and eosin (H&E)-stained esophageal biopsy tissue section. B) Overlay of the original H&E-stained image generated by the digital pathology image analysis algorithm, used for automated analysis.

Figure 2. A concordance analysis (sensitivity assessment) was conducted between manually (left, green labels) and automated (right, blue labels) eosinophil count in the same high-power field (HPF) that was selected by a pathologist.

Validation of the algorithm performance involved evaluating both specificity and sensitivity of the algorithm to detect the object of interest. A board-certified pathologist visually confirmed the algorithm’s specificity in detecting eosinophils within the epithelial tissue compartment and its generation of a density-based heatmap. To assess sensitivity, the concordance between the pathologist’s manual and the automated PEC within the same HPF was calculated. The algorithm demonstrated at least 90% specificity for eosinophil detection and excellent concordance with manual counts (C-index=0.916), resulting in a 96% precision rate in identifying samples that meet the diagnostic threshold for EoE (≥15 PEC/HPF).

Empowering pathologists with automated PEC analysis

The goal of developing this augmented reading tool is to enhance, not replace, the role of pathologists. By automating the time-consuming task of quantifying PEC in esophageal biopsies, this tool, when available, will aim to improve the accuracy and efficiency of EoE histological assessment. This allows pathologists to dedicate more time to evaluating other features within the Eosinophilic Esophagitis Histologic Scoring System (EoEHSS), which can be particularly valuable for EoE histology characterization.

Alimentiv remains at the forefront of imaging solutions for gastrointestinal clinical trials. Building on this commitment, we are actively integrating the PEC digital image analysis algorithm into Lucidity, our proprietary histopathology imaging platform. The goal of this integration is to soon provide pathologists with an augmented reading tool for conducting their assessment of EoE biopsies in clinical trials.

Contact us to learn more about our AI-based algorithm and how we can support your GI clinical trials.

References

  1. Khan S, Guo X, Liu T, et al. An Update on Eosinophilic Esophagitis: Etiological Factors, Coexisting Diseases, and Complications. Digestion. 2021;102(3):342-356. doi:10.1159/000508191
  2. Gonsalves N, Policarpio-Nicolas M, Zhang Q, Rao MS, Hirano I. Histopathologic variability and endoscopic correlates in adults with eosinophilic esophagitis. Gastrointest Endosc. 2006;64(3):313-319. doi:10.1016/j.gie.2006.04.037
  3. O’Shea KM, Rochman M, Shoda T, Zimmermann N, Caldwell J, Rothenberg ME. Eosinophilic esophagitis with extremely high esophageal eosinophil counts. J Allergy Clin Immunol. 2021;147(1):409-412.e5. doi:10.1016/j.jaci.2020.05.045
  4. Odetola S, Feulefack J, Sergi CM. Eosinophilic esophagitis: absolute eosinophilic count, peak eosinophilic count, and potential biomarkers of eosinophilic degranulation products-an in-depth systematic review. Transl Pediatr. 2024;13(3):474-483. doi:10.21037/tp-23-478
  5. Turner KO, Collins MH, Walker MM, Genta RM. Quantification of Mucosal Eosinophils for the Histopathologic Diagnosis of Eosinophilic Gastritis and Duodenitis: A Primer for Practicing Pathologists. Am J Surg Pathol. 2022;46(4):557-566. doi:10.1097/PAS.0000000000001843
  6. Lefevre P, Guizzetti L, McKee TD, et al. Development and Validation of a Digital Analysis Method to Quantify CD3-immunostained T Lymphocytes in Whole Slide Images of Crohn’s Disease Biopsies. Appl Immunohistochem Mol Morphol. 2022;30(7):486-492. doi:10.1097/PAI.0000000000001035
  7. Lefevre, P. et al. (2024) ‘961 development of a digital pathology analysis algorithm to augment quantification of peak eosinophil count in biopsies from patients with eosinophilic esophagitis’, Gastroenterology, 166(5). doi:10.1016/s0016-5085(24)01010-2