Efficient disease detection in gastrointestinal videos – global features versus neural networks

• DOI: 10.1007/s11042-017-4989-y
• Publication Date: 2017-07-19T17:32:17Z
• Authors: Konstantin Pogorelov; Michael Riegler; Sigrun Losada Esk...; Thomas de Lange; Dag Johansen; Carsten Griwodz; Peter Thelin Schmidt; Pål Halvorsen
• Keywords: 02 engineering and technology; Deep learning neural networks; 3. Good health; VDP::Teknologi: 500::Medisinsk teknologi: 620; Medical; Automatic disease detection; Global and local image features; Performance evaluation; 0202 electrical engineering, electronic engineering, information engineering; Information retrieval; Algorithmic screening; VDP::Technology: 500::Medical technology: 620

Analysis of medical videos from the human gastrointestinal (GI) tract for detection and localization abnormalities like lesions diseases requires both high precision recall. Additionally, it is important to support efficient, real-time processing live feedback during (i) standard colonoscopies (ii) scalability massive population-based screening, which we conjecture can be done using a wireless video capsule endoscope (camera-pill). Existing related work in this field does neither provide necessary combination accuracy performance detecting multiple classes simultaneously nor particular disease tasks. In paper, complete end-to-end multimedia system presented where aim tackle automatic analysis GI videos. The includes an entire pipeline ranging data collection, analysis, visualization. combines deep learning neural networks, information retrieval, global local image features order implement multi-class classification, localization. Furthermore, built modular way, so that easily extended deal with other types abnormalities. Simultaneously, developed efficient doctors reasons when potentially applied algorithmic screenings future. Initial experiments show our has polyp at least as good state-of-the-art systems, provides additional novelty terms performance, low resource consumption ability extend new diseases.