AIQEN – Artificial Intelligence for Quality in Endoscopy
Short description
Endoscopy is an examination of the gastrointestinal tract. Gastroscopy and colonoscopy are most common. During gastroscopy, the mucosa of the oesophagus, stomach and duodenum is examined, while during colonoscopy, the rectum and colon are examined. During the examination, it is easy to miss abnormalities because the intestine has many folds and is constantly moving. The examinations can also be experienced as intrusive and painful.
We research development and evaluation of AI software. The purpose is to help the endoscopist detect all mucosal changes.
We also research AI for capsule endoscopy (camera pill). Hopefully, the method will replace many gastroscopies and colonoscopies, which means that patients will experience less discomfort from the examinations and that the healthcare system will save resources.
About AIQEN research group - Artificial Intelligence for Quality in Endoscopy
The research group aims to improve the quality of gastrointestinal endoscopy through the development, evaluation and implementation of artificial Intelligence (deep learning) based image analysis. As a consequence, the most intrusive endoscopies can partially be replaced by the less invasive capsule endoscopy and improve patient integrity and experience.
At this stage we focus on three projects.
- CAI-PD (Comparison of Artificial Intelligence systems for Polyp Detection). The study evaluates whether or not polyp detection rates are different between commercially available CADe systems for polyp detection.
PI: Jonas Varkey
- AIQEN-CPC (Artificial Intelligence for Improved Quality in ENdoscopy for Colorectal cancer Prevention in Colitis) consists of three parts:
- a. MST-C (Multiple Staining Techniques in Colitis) to assess whether it is possible to replace time consuming chromoendoscopy (manual staining) for dysplasia detection by digital staining during colonoscopy.
- b. CID-AID (Colitis Image Database for Artificial Intelligence software Development), aims to create a large, labelled video database of IBD surveillance colonoscopies.
- c. AIDDUC- (Artificial Intelligence for Detection of Dysplasia in Ulcerative Colitis) aims to develop AI algorithms for automatic detection of dysplasia in ulcerative colitis and subsequently reduce the incidence of colorectal cancer in this group of patients.
PI: Thomas de Lange
- AIQEN-IDA- Artificial Intelligence for Capsule Endoscopy in Iron Deficiency Anaemia consists of three parts:
- a. MIP-IDA (Mini-Invasive endoscopy for Pan-intestinal diagnosis of Iron Deficiency Anemia) assesses whether it is possible to replace invasive gastroscopy and colonoscopy with non-invasive capsule endoscopy for diagnostic work-up in IDA.
- b. IDA-AID (Iron Deficiency Anemia image & video database for AI software Development) target to create a comprehensive open-access video capsule endoscopy database for AI software development of lesions detected in IDA.
- c. AID-MPI (AI software Development for Mini-invasive Pan-intestinal diagnosis of Iron deficiency anemia) aims to create a performant AI software for detection of bleeding sources in the entire bowel.
A successful project may substantially reduce the burden on patients.
PI: Nikolaos Papachrysos
Key publications
- PolypConnect: Image inpainting for generating realistic gastrointestinal tract images with polyps.
- SinGAN-Seg: Synthetic training data generation for medical image segmentation
- Visual explanations for polyp detection: How medical doctors assess intrinsic versus extrinsic explanations
- Assessing generalisability of deep learning-based polyp detection and segmentation methods through a computer vision challenge
- MedAI: Transparency in Medical Image Segmentation
- Kvasir-instrument: Diagnostic and therapeutic tool segmentation dataset in gastrointestinal endoscopy
- PolypGen: A multi-center polyp detection and segmentation dataset for generalisability assessment
- Nanonet: Real-time polyp segmentation in video capsule endoscopy and colonoscopy
- Kvasir-Capsule, a video capsule endoscopy dataset
- DeepSynthBody: the beginning of the end for data deficiency in medicine
- A comprehensive analysis of classification methods in gastrointestinal endoscopy imaging
- A comprehensive study on colorectal polyp segmentation with ResUNet++, conditional random field and test-time augmentation
- HyperKvasir, a comprehensive multi-class image and video dataset for gastrointestinal endoscopy
- Kvasir-seg: A segmented polyp dataset
- Resunet++: An advanced architecture for medical image segmentation
