MultiModal Endoscopy for Inflammatory Lesions of the Stomach – EMMIE

The project is to develop an innovative multimodal technology acquisition and multispectral imaging for early detection and characterization of mucosal inflammation preferentially occurring in two areas of the stomach, distal portion (antrum) and the junction with the esophagus (cardia). These inflammatory lesions can induce various diseases and are undetectable under white light examination, forcing the gastroenterologist to perform systematic and non-oriented biopsies of the mucosa. Furthermore, the findings of the endoscopic examination are strongly dependent on the experience of the operator and the procedure conditions, compromising the reproducibility of the results. While several studies have shown that the diffuse reflectance properties of the gastric mucosa vary between healthy tissue and inflammatory lesions, few systems are using optical imaging in endoscopy and they are limited to the acquisition or estimating a very small number of spectral bands. The detection of inflammatory zones can be improved using specific wavelengths characteristic of lesions. These ascertainments justify the development of multispectral imaging techniques for early diagnosis and characterization of inflammation of the stomach mucosa by a reproducible and quantitative technique for tracking the results.
However, the implementation of this new imaging modality is very challenging in an endoscopic context. It requires several innovations in both the instrumental aspect and the processing and analysis of images and signals. We propose to develop a technologically innovative prototype imager enabling simultaneous acquisition of a video composed of multispectral images and a colour video in white light through a fiberoptic connected to a multispectral camera that will be inserted into the channel operator of a standard gastroscope. We intend to establish the relationships between the grades of inflammation and the modification of the tissue reflectance using ex-vivo analysis in an animal model and in-vitro analysis via explant culture. Concerning the image analysis, besides the development of robust multimodal registration, we wish to design real time algorithms that provide a diagnostic aid during endoscopy procedures alerting the physician when the analysis of spectral images detects potentially inflammatory areas. The detection and characterization of these lesions will be postoperatively refined based on the estimation of biophysical parameters of inflammatory lesions based on a light-tissue interaction model. Then, we will develop mosaicing algorithms that map video sequences recorded during a gastroscopy providing the spectral information on a wide field of view of an area of interest. This will improve the traceability of examinations.
Our approach is completely within the framework of the challenge "Life, Health and Wellbeing". It is the link between the theoretical research in image processing and a first implementation in clinical practice thanks to a coherent distribution of skills and tasks between the six partners. In addition to the scientific communications, technology transfer is considered with the realization of a prototype at the end of the project using the result of a market research.