Artificial intelligence (AI) is predicted to have profound effects on the future of video capsule endoscopy (VCE) technology. The potential lies in improving anomaly detection while reducing manual labour. Existing work demonstrates the promising benefits of AI-based computer-assisted diagnosis systems for VCE. They also show great potential for improvements to achieve even better results. Also, medical data is often sparse and unavailable to the research community, and qualified medical personnel rarely have time for the tedious labelling work. We present Kvasir-Capsule, a large VCE dataset collected from examinations at a Norwegian Hospital. Kvasir-Capsule consists of 117 videos which can be used to extract a total of 4,741,504 image frames. We have labelled and medically verified 47,238 frames with a bounding box around findings from 14 different classes. In addition to these labelled images, there are 4,694,266 unlabelled frames included in the dataset. The Kvasir-Capsule dataset can play a valuable role in developing better algorithms in order to reach true potential of VCE technology.

Artificial intelligence (AI) is predicted to have profound effects on the future of video capsule endoscopy (VCE) technology.The potential lies in improving anomaly detection while reducing manual labour. However, medical data is often sparse andunavailable to the research community, and qualified medical personnel rarely have time for the tedious labelling work. Inthis respect, we presentKvasir-Capsule, a large VCE dataset collected from examinations at Bærum Hospital in Norway.Kvasir-Capsuleconsists of118videos from which we can generate2,830,089image frames. We have labelled and medicallyverified44,260frames with a bounding box around detected anomalies from 13 different classes of findings. In addition to theselabelled images, there are2,785,829unlabelled frames included in the dataset. Initial experiments demonstrate the potentialbenefits of AI-based computer-assisted diagnosis systems for VCE. However, they also show that there is great potentialfor improvements, and theKvasir-Capsuledataset can play a valuable role in developing better algorithms in order for VCEtechnology to reach its true potential

A system for robotically assisted retinal surgery has been developed to rapidly and safely place lesions on the retina for photocoagulation therapy. This system provides real-time, motion stabilized lesion placement for typical irradiation times of 100 ms. The system consists of three main subsystems: a digital-based global tracking subsystem; a fast, analog local tracking subsystem; and a confocal reflectance subsystem to control lesion parameters dynamically. We have reported previously on these individual subsystems. This paper concentrates on the development of a second hybrid system prototype. Considerable progress has been made toward reducing the footprint of the optical system, simplifying the user interface, fully characterizing the analog tracking system, using measurable lesion reflectance parameters to develop a noninvasive method to infer lesion depth, and integrating the subsystems into a seamless hybrid system. These system improvements and progress toward a clinically significant system are covered in detail within this paper. The tracking algorithms and concepts developed for this project have considerable potential for application in many other areas of biomedical engineering.

A tracking algorithm has been developed to efficiently track a moving object in two-dimensional image space. The algorithm employs a limited exhaustive template matching scheme that combines the accuracy of an exhaustive search with the computational efficiency of a coarse-fine template matching scheme. The overall result is an accurate, time-efficient tracking algorithm. After providing a theoretical discussion of the algorithm, three separate biomedical applications of the algorithm are described: (1) stabilizing an irradiating laser on the retinal surface for photocoagulation treatment, (2) measuring target fixation eye movement to construct pattern densities at the retina, and (3) tracking a rat swimming in a Morris Water Maze for psychophysiological studies. Results for each application is provided. The paper concludes with a discussion of the relative merits of the tracking algorithm and recommendations for methods to improve the performance of the algorithm. © 2001 SPIE and IS&T.

A system for robotically assisted retinal surgery has been developed to rapidly and safely place lesions on the retina for photocoagulation therapy. This system provides real-time, motion stabilized lesion placement for typical irradiation times of 100 ms. The system consists ofthree main subsystems: a global, digital-based tracking subsystem; a fast, local analog tracking subsystem; and a confocal reflectance subsystem to control lesion parameters dynamically. We have reported on these subsystems in previous SPIE presentations. This paper concentrates on the development ofthe second hybrid system prototype. Considerable progress has been made toward reducing the footprint ofthe optical system, simplifying the user interface, fully characterizing the analog tracking system and using measurable lesion reflectance growth parameters to develop a noninvasive method to infer lesion depth. This method will allow dynamic control oflaser dosimetry to provide similar lesions across the non-uniform retinal surface. These system improvements and progress toward a clinically significant system are covered in detail within this paper.

Colorectal cancer is a severe health issue globally and a significant cause of cancer-related mortality, but it is treatable if found at an early stage. Early detection is usually done through a colonoscopy, where clinicians search for cancer precursors called polyps. Research has shown that clinicians miss between 14% and 30% of polyps during standard screenings of the gastrointestinal tract. Furthermore, once the polyps have been found, clinicians often overestimate the size of the polyps. In this respect, automatic analysis of medical images for detecting and locating polyps is a research area where machine learning has excelled in recent years. Still, current models have much room for improvement. In this paper, we propose a novel approach based on learning to segment within several grids, which we introduce to U-Net and Pix2Pix architectures. In short, we have experimented using several grid sizes, and using two opensource polyp segmentation datasets for cross-data training and testing. Our results suggest that segmentation at lower resolutions produces better results at the cost of less precision, which proved useful for the cases where higher precision segmentations gave limited results. Generally, compared to traditional U-Net and Pix2Pix, our grid-based approaches improve segmentation performance.