APPLICATION OF ARTIFICIAL INTELLIGENCE IN MEDICAL IMAGE RECOGNITION

Abstract

The article is devoted to the study of image recognition in medicine using deep learning. The use of deep learning allows automating image processing and analysis, which significantly reduces the human factor and increases the accuracy of diagnoses. Artificial intelligence, in particular deep learning, is actively used to recognize abnormalities in X-rays, ultrasound images, MRI, and CT scans, so this technology is actively developing in the areas of oncology and cardiology. The algorithms can learn from large amounts of data, which allows them to identify patterns that may be unnoticed or unclear to human observation, which is made possible by convolutional layers that use filters to detect local features in images. The purpose of the publication is to study the current state of the art of medical image recognition and summarize the latest research in this area. Methodology. The article reviews and analyzes the literature on the use of deep learning, its advantages, disadvantages and limitations relative to traditional methods in image recognition, considers the necessary steps for building image recognition systems, and proves the importance of convolutional neural networks (CNNs). Conclusions. Although medical image recognition is not the most popular area of application of convolutional neural networks today, it is very important for providing more effective treatment to the population. The most researched and relevant areas in medicine are lungs, heart, breast, liver, histology, and eyes. Modern publications of scientists have proven the high accuracy of convolutional neural networks in diagnosing diseases. However, some indicators can still be significantly improved, which gives room for further research. Convolutional neural networks have demonstrated high accuracy in recognizing patterns in medical images, which can contribute to early diagnosis of diseases in the above-mentioned medical fields. In addition, the use of CNNs will help automate processes, which will reduce the workload of medical staff for more complex cases; increase the efficiency of processing large amounts of data; and reduce the number of erroneous image interpretations. Today, convolutional neural networks can be called an assistant for medical staff, but a lot of research and investment is still needed for their widespread use by doctors, but if these conditions are met, their potential is worth further research.