Provided are an image processing device, an endoscope system, an image processing method, and a program capable of automatically differentiating a medical image including a scene of interest and supporting saving of the medical image according to a differentiation result. The image processing device includes a medical image acquisition unit (41) that acquires a medical image, a scene-of-interest recognition unit (51) that recognizes a scene of interest from the medical image acquired using the medical image acquisition unit, a degree-of-similarity calculation unit (52) that, for the scene of interest recognized using the scene-of-interest recognition unit, calculates a degree of similarity between the medical image acquired using the medical image acquisition unit and a standard image determined for the scene of interest, and a saving processing unit (53) that executes processing for saving the medical image in a saving device based on the degree of similarity calculated using the degree-of-similarity calculation unit.

Provided is a disease diagnosis support method employing endoscopic images of a digestive organ using a neural network, and the like. The disease diagnosis support method employing endoscopic images of a digestive organ using a neural network trains the neural network by using first endoscopic images of the digestive organ, and corresponding to the first endoscopic images, at least one of definitive diagnosis result of being positive or negative for the disease of the digestive organ, a past disease, a severity level, and information corresponding to an imaged region. The trained neural network outputs, based on second endoscopic images of the digestive organ, at least one of a probability of being positive and/or negative for the disease of the digestive organ, a probability of a past disease, a severity level of the disease, and the information corresponding to the imaged region.

An image processing system connected to an endoscope and processing in real-time endoscopic images to identify suspicious tissues such as polyps or cancer. The system applies preprocessing tools to clean the received images and then applies in parallel a plurality of detectors both conventional detectors and models of supervised machine learning-based detectors. A post processing is also applied in order select the regions which are most probable to be suspicious among the detected regions. Frames identified as showing suspicious tissues can be marked on an output video display. Optionally, the size, type and boundaries of the suspected tissue can also be identified and marked.

The present disclosure relates to systems and methods for processing real-time video and detecting objects in the video. In one implementation, a system is provided that includes an input port for receiving real-time video obtained from a medical image device, a first bus for transferring the received real-time video, and at least one processor configured to receive the real-time video from the first bus, perform object detection by applying a trained neural network on frames of the received real-time video, and overlay a border indicating a location of at least one detected object in the frames. The system also includes a second bus for receiving the video with the overlaid border, an output port for outputting the video with the overlaid border from the second bus to an external display, and a third bus for directly transmitting the received real-time video to the output port.

Image recognition may include obtaining a first image, segmenting the first image into a plurality of first regions by using a target model, and searching for a target region among bounding boxes in the first image that use points in the first regions as centers. The target region is a bounding box in the first image in which a target object is located. The target model is a pre-trained neural network model configured to recognize from an image, a region in which the target object is located. The target model is obtained through training by using positive samples with a region in which the target object is located marked and negative samples with a region in which a noise is located marked. The target region is marked in the first image to improve accuracy of target object detection in an image.

The present disclosure is directed towards systems and methods that leverage machine-learned models to decrease the rate at which abnormal sites are missed during a gastroenterological procedure. In particular, the system and methods of the present disclosure can use machine-learning techniques to determine the coverage rate achieved during a gastroenterological procedure. Measuring the coverage rate of the gastroenterological procedure can allow medical professionals to be alerted when the coverage output is deficient and thus allow an additional coverage to be achieved and as a result increase in the detection rate for abnormal sites (e.g., adenoma, polyp, lesion, tumor, etc.) during the gastroenterological procedure.

A method of processing images captured using an endoscope comprising a camera is disclosed. According to this method, regular images captured by the camera are received while the endoscope is maneuvered by an operator to travel through a human gastrointestinal (GI) tract. The regular images are mosaicked to determine any missed or insufficiently imaged area in a section of the human GI tract already travelled by the endoscope. If any missed or insufficiently imaged area is detected, information regarding any missed or insufficiently imaged area is provided to the operator.

A method for performing real-time detection and displaying of polyps in optical colonoscopy, includes a) acquiring and displaying a plurality of real-time images within colon regions to a video stream frame rate, each real-time image comprising a plurality of color channels; b) selecting one single color channel per real-time image for obtaining single color pixels; c) scanning the single color pixels across each the real-time image with a sliding sub-window; d) for each position of the sliding sub-window, extracting a plurality of single color pixels local features of the real-time image; e) passing the extracted single color pixels local features of the real-time image through a classifier to determine if a polyp is present within the sliding sub-window; f) real-time framing on display of colon regions corresponding to positions of the sliding sub-window wherein polyps are detected. A system for carrying out such a method is also provided.

Embodiments of this application include an image processing method and apparatus, a non-transitory computer-readable storage medium, and an electronic device. In the image processing method a to-be-predicted medical image is input into a multi-task deep convolutional neural network model. The multi-task deep convolutional neural network model includes an image input layer, a shared layer, and n parallel task output layers. One or more lesion property prediction results of the to-be-predicted medical image is output through one or more of the n task output layers. The multi-task deep convolutional neural network model is trained with n types of medical image training sets, n being a positive integer that is greater than or equal to 2.

The present disclosure relates to computer-implemented systems and methods for training and using generative adversarial networks. In one implementation, a system for training a generative adversarial network may include at least one processor that may provide a first plurality of images including representations of a feature-of-interest and indicators of locations of the feature-of-interest and use the first plurality and indicators to train an object detection network. Further, the processor(s) may provide a second plurality of images including representations of the feature-of-interest, and apply the trained object detection network to the second plurality to produce a plurality of detections of the feature-of-interest. Additionally, the processor(s) may provide manually set verifications of true positives and false positives with respect to the plurality of detections, use the verifications to train a generative adversarial network, and retrain the generative adversarial network using at least one further set of images, further detections, and further manually set verifications.