Human-machine collaboration: Improving endoscopic detection and characterization of colorectal neoplasia

One of the most prominent clinical challenges in colorectal cancer (CRC) prevention is to further optimize the endoscopic detection and characterization (optical diagnosis) of colorectal polyps. Three approaches to improve this are investigated in this thesis. The first part of the thesis (chapters 2-5) focuses on advanced endoscopic techniques. In this part we describe, amongst others, the results of an international multicenter randomized controlled trial comparing Linked Color imaging (LCI) with high-definition white-light endoscopy (HD-WLE) for polyp detection during surveillance of individuals with Lynch syndrome. Another chapter evaluates the additional diagnostic value of dye-based chromoendoscopy compared to standard-definition WLE and HD-WLE for surveillance of individuals with Lynch syndrome. The second part of the thesis (chapters 6-9) focuses on the development of dedicated training programs and competence standards. In collaboration with the European Society for Gastrointestinal Endoscopy (ESGE) a training curriculum for optical diagnosis practice across Europe has been developed. Furthermore, competence standards for the optical diagnosis strategy of diminutive colorectal polyps were developed, based on a simulation study and a Delphi consensus procedure. The third part of the thesis (chapters 10-12) focuses on the use of machine learning techniques for the endoscopic detection and characterization of colorectal neoplasia. These techniques, often referred to as computer aided detection (CADe) and computer aided diagnosis (CADx) can assists endoscopist during colonoscopy, by discriminating between the different types of polyps or by indicating that polyps are missed. In chapter 13 the main finding of this thesis are reviewed and future research directions are discussed