A method for controlling a system that assists an endoscopy diagnosis based on artificial intelligence, the method comprising:
- providing an endoscopic image for internal organs of a patient;
- receiving medical information of the patient;
- analyzing, using an image processing program, the endoscopic image to detect lesion information including a size, a shape, and coordinates of at least one lesion, matching the detected lesion information and the medical information, generating, using a lesion diagnosis program, at least one diagnosis information of malignancy and a malignancy probability corresponding to a result of the matching, and outputting a notification signal according to the detected lesion information and the at least one diagnosis information;
- and
- displaying the at least one diagnosis information on a screen according to the notification signal,
- wherein the endoscopic image is matched and classified with respective regions of the internal organs of patient, respectively, using the image processing program to which a deep learning model is applied, search information for distinguishing a search completion region and an unsearched region is generated based on the endoscopic image, and the notification signal is output according to the generated search information,
- wherein the deep learning model of the image processing program is pre-learned by first acquiring a plurality of frame images in which lesions appear in a plurality of endoscopic images acquired by pre-capturing internal organs of multiple patients,
- wherein the lesion diagnosis program is a deep learning-based program that is pre-matched and learned by first acquiring the plurality of frame images in which the lesions appear and a plurality of medical information of the multiple patients corresponding thereto,
- wherein the endoscopic image is acquired at a plurality of frames per second,
- wherein the plurality of frame images is converted to meet a condition of the image processing program, and
- wherein the plurality of frame images in which the lesions appear in the plurality of endoscopic images acquired by pre-capturing the internal organs of the multiple patients, the plurality of medical information for the multiple patients corresponding to the plurality of frame images in which the lesions appear, and a learning environment of the image processing program and a learning environment of the lesion diagnosis program are stored in a database.