Leveraging a hybrid AI modeling technique in detection of Helicobacter pylori in gastric biopsies

Helicobacter pylori is a major causative factor in gastric carcinoma, making detection in gastric biopsies a matter of great interest and need. Traditional diagnostic techniques such as endoscopy-guided gastric biopsy, rapid urease test, polymerase chain reaction, or tissue culture often face turnaround time and accuracy challenges.

This study explores the development of advanced hybrid AI modeling using deep learning algorithms to enhance precision in detecting HP in histological specimens. Retrospectively collected images from endoscopic biopsies, stained with Giemsa and hematoxylin and eosin, were analyzed and marked by a pathologist to classify HP as positive or negative cases. The study used 1395 images from a medical college hospital. A hybrid AI model was created to improve classification accuracy.

Performance metrics for the hybrid model revealed exceptional results: an overall accuracy of 99.6%, sensitivity of 99.3%, specificity of 100%, precision of 100%, and F1-score of 99.7%. The receiver operating characteristic curve analysis showed an Area Under the Curve of 1.00, indicating perfect classification ability.

This paper highlights the opportunities of AI technologies to simplify and enhance the diagnostic workflow in busy laboratories, in turn, to help clinical decision-makers decide to manage HP accurately and effectively.