Investigation of Usability Of Artificial Intelligence Semantic Video Processing Methods in Medicine

Aim: The goal of this study is to produce user-friendly software for healthcare professionals with various approaches such as detection, identification, classification, and tracking polyps contained in endoscopic images utilizing appropriate video/image processing techniques CNN architecture. 
 Material Method: There were 345 photos total the study. These photographs are depicting anatomical milestones, clinical findings, or gastrointestinal procedures digestive tract that have been documented validated by medical specialists (skilled endoscopists). Each class has hundreds images. downloaded from https://datasets.simula.no/kvasir, which a free source educational research purposes. In modeling phase, Max-Margin object detection technique (MMOD), one deep neural network designs Dlib package, employed. data set was separated 80% training 20% test dataset using simple cross-validation method (hold-out). Precision, recall, F1-score, average precision (AP), mean (mAP), ideal localization recall (oLRP), optimal LRP (moLRP), intersection over union (IoU) used evaluate model performance.
 Results: When previously described steps performed on open-access video image current study, all performance metrics examined received value 1, whereas, precision, sensitivity, AP, mAP, oLRP, moLRP 98%, 90%, 94%, 89%, 48%, 48% respectively.
 Conclusion: proposed approach found make accurate predictions diagnosis based values calculated criteria.