Mycophenolate mofetil-induced colitis versus colonic graft-versus-host disease: a comparative histologic study with...

Area of Science:

• Gastroenterology
• Hematology
• Pathology

Background:

• Distinguishing Mycophenolate Mofetil (MMF)-induced colitis from Graft-versus-Host Disease (GVHD) is crucial in transplant patients.
• Histopathologic features are key, but objective quantification of specific cell types like eosinophils can be challenging.

Purpose of the Study:

• To compare the histopathologic features of MMF-induced colitis and colonic GVHD.
• To develop and validate a deep learning model for semi-automated eosinophil quantification in the colon.

Main Methods:

• Retrospective review of MMF-induced colitis and GVHD cases.
• Histologic features were analyzed, including apoptotic counts and neuroendocrine aggregates.
• A deep learning convolutional neural network (CNN) was trained to quantify lamina propria eosinophils.

Main Results:

• GVHD cases showed significantly higher apoptotic counts compared to MMF-induced colitis.
• Neuroendocrine aggregates were observed in GVHD but not in MMF-induced colitis.
• The deep learning CNN identified significantly higher eosinophil counts in MMF-induced colitis compared to GVHD.

Conclusions:

• Lower eosinophil counts, higher apoptotic counts, and presence of neuroendocrine aggregates favor GVHD.
• AI-powered eosinophil quantification can serve as an ancillary tool for differential diagnosis.
• A high threshold for diagnosing MMF-induced colitis is recommended in stem cell transplant patients due to its rarity.