Automatic Polyp Detection in Endoscope Images Using a Hessian Filter

An endoscope is a medical instrument that acquires images inside the human body. This paper proposes a new approach for the automatic detection of polyp regions in an endoscope image using a Hessian filter and machine learning techniques. Previous approaches tried to detect candidate polyp regions based on rectangular patches. But, a purely patch-based approach can miss classify candidate regions because other information necessarily is included in each rectangular patch. Here, a Hessian filter is used to detect image regions corresponding to blob-like structures. Detailed color and edge features are extracted only for the detected candidate regions. SVMs (with Boosting) are constructed to classify candidate regions as polyps. The new approach is demonstrated experimentally. High accuracy is achieved.