Enhancing Gastrointestinal Diagnosis: CBAM-Integrated CNN for High-Precision Multi-Class Anatomical Landmark Identification

Gastrointestinal diseases are highly prevalent in humans, posing a significant healthcare concern. Numerous disorders within the gastrointestinal tract, such as obscure gastrointestinal bleeding, ulcer infections, and both benign and malignant tumors, can manifest without obvious symptoms. If left untreated, these conditions have the potential to progress and ulti-mately lead to gastrointestinal cancer. Conventionally, endoscopic examination is employed to assess these gastrointestinal diseases, with anatomical landmarks featuring characteristics that guide gastroenterologists during endoscopy and assist experts in ensuring a thorough examination. Automated detection of anatomical anomalies in endoscopic images is essential for aiding clinical assessments and streamlining the expenditure and duration associated with screening the gastrointestinal tract. This research introduced an innovative Convolutional Block Attention Module (CBAM) based deep learning method to identify anatomical landmarks in the gastrointestinal (GI) tract. This recently devised architectural framework was tested on the anatomical landmarks of the Kvasir-V2dataset. The incorporation of CBAM with CNN in this research yielded a paradigm-shifting enhancement, significantly elevating accuracy and efficiency in identifying intricate gastrointestinal anatomical landmarks. The suggested method successfully identified anatomical anomalies with an impressive 99.1% accuracy. The proposed CBAM-based deep learning approach outperformed most current studies in gas-trointestinal classification, highlighting remarkable superiority in both accuracy and efficiency.