Improved Pathology Detection in Wireless Capsule Endoscopy Images through Artificial Intelligence and 3D Reconstruction

Prosjektet "Improved Pathology Detection in Wireless Capsule Endoscopy Images through Artificial Intelligence and 3D Reconstruction" har som mål å på en effektiv måte forbedre nøyaktigheten for deteksjon av tarmsykdommer gjennom bruk av kapselendoskopi. Dette vil føre til bedre behandling. Nøyaktige og effektive verktøy for å stille diagnose er essensielt for tilrettelegging av masseundersøkelser og oppfølging for store deler av befolkningen. Ved å stille diagnose(r) tidlig i sykdomsforløpet minsker man sannsynligheten for alvorlig sykdomsforløp. Prosjektet har fokusert på to metoder: den første er å utvikle mer nøyaktige og effektive automatiske patologi deteksjon algoritmer gjennom bruk av kunstig intelligens, hvor man fokuserer på bruk av semi-veiledet og ikke-veiledet læring. Dette har resultert i dyp læringsteknikker som har forbedret diagnose av patologier i kapselendoskopi. Den andre har vært å lage 3D modeller av tarmen som vil forenkle gjenkjenning av patologier i kapselendoskopi bilder, dette er basert på struktur fra bevegelse. Prosjektet har også resultert i et større datasett som er åpent tilgjengelig. Prosjektet har hatt fire partnere; Sykehuset Innlandet (Norge), University of Thessaly (Hellas), Sheffield Teaching Hospitals (England) and Max Planck Institute for Informatics (Tyskland), alle er anerkjente på sine felt og bringer komplementære ferdigheter til prosjektet. Dette samarbeidet har økt verdiskapningen gjennom bruk av forskningen i praksis.

The project has advanced computer aided diagnosis in the face of data and label adversity, proposed techniques to generate synthetic data and evaluate its quality with a view to discovering factors that makeup wireless capsule endoscopy data; and explore the automation of explanation generation for diagnosis. The proposed methodologies address some of the key challenges faced in diagnosing wireless capsule endoscopy pathologies, including the scarcity of labeled data and the difficulty of interpreting complex images, both of which could have significant implications for improving diagnostic accuracy and ultimately patient outcomes. The project has also suggested techniques for 3D shape reconstruction from capsule endoscopy video and evaluation of these have found them to be very useful by gastroenterologists. The results show that the use of 3D reconstructed models, in addition to images, are found to be highly beneficial, and in some cases, the gastroenterologists even preferred 3D models over the original images.

As indicated by its title: "Improved Pathology Detection in Wireless Capsule Endoscopy Images through artificial Intelligence and 3D Reconstruction", the core concept of the project is to facilitate more accurate and efficient detection of intestinal diseases through capsule endoscopy, leading to better healthcare. The project is highly relevant and of great benefit to society. Accurate and efficient tools for diagnosis are essential in order to facilitate intestinal screening and monitoring programs for large parts of the population that seek to detect diseases at an early stage and thereby prevent severe diseases from developing. The project is divided into three work packages. The first focuses on developing more accurate and efficient automatic pathology detection algorithms than state-of-the art methods. These will also be compared to the performance of trained gastroenterologists. The second work package focuses on constructing 3D models based on wireless capsule endoscopy images that will help gastroenterologists more easily detect pathologies in the intestinal wall. The third work package is dedicated to the combination of automatic pathology detection and 3D reconstruction. The goal is twofold: i) automatically construct 3D model based on pathology detection algorithms predictions for enhanced viewing by gastroenterologists. ii) Improve automatic pathology detection using 3D data as input. The project includes one national (Innlandet Hospital) and three international partners (University of Thessaly, Sheffield Teaching Hospitals and Max Planck Institute for Informatics), whom of each is renowned in their field and highly capable of bringing complementary skills. This collaboration facilitates and enhances the value creation through the utilization of the research results in practice. It also ensures that the results are brought into practice faster and thereby benefit society.