Automatic Detection of Gastrointestinal Tract Lesions
GastroIntestinal Track diagnoser (GITdiagnoser) aims at the development of a Computer-aided Diagnosis (CAD) system for aid in the diagnosis of gastrointestinal track (GIT) lesions in clinical based on videos of wireless capsule endoscopy (VCE). The main output is CAD prototype to be evaluated in the Gastroenterology service of the Centro Hospitalar Universitário do Porto (CHUP), Porto.
Main Task Goals:
Private Dataset Creation
To create a wide private and anonymized dataset of wireless endoscopic videos with medical annotations.
Informative Regions Identification
Classify the pixels of the VCE images as informative or non-informative to create a map of the informational regions of the image.
Classify the VCE images according to the part of the GI tract (esophagus, stomach, small intestine, and large intestine), and estimates the displacement between images.
Detect candidates for abnormalities, i.e., different occurrences from that which appears in tests considered normal to include pathologies of the GI tract and classifies them as abnormalities or non-abnormalities.
Classification of the detected abnormalities in the various types of known lesions/pathologies or as an unknown abnormality.
Producing a user-friendly interface that will allow the reviewer to navigate through the VCE, to view detected anomalies and to annotate the VCE that can later added to the database.