Deep learning and image recognition in various application fields, development toward that. Medicine and sports science, but also in the field of image recognition and the advancement of essential technologies. However, analysis time or error in the image recognition system to introduce the challenges have been. In the meantime, Nara Institute of Science and technology is 13, the 3-dimensional CT images from deep learning, using a highly accurate recognition system was developed and announced.
Nara Institute of Science and technology has developed image recognition systems,3-dimensional CT images from musculoskeletal system constituting the muscles and bones individually, quickly and with high precision recognition system. The configuration that muscle is 19 kinds of bones, 3 types and for image recognition is a lot of data using the deep learning that was required to do. Deep learning by accumulating the data,automatically learning and image recognition error and computation time radically simplify computing for millions of users, delivering applications as an on-demand or.
Specifically, the image recognition of the error is 1 mm or less, the calculation time is from the pelvis to the knee of CT images in 5 minutes, and practical enough to endure to get to the level reached. However, various applications can be applied to a system based significance is greater.
Also this research results of features, the recognition results for the”confidence”to output at the same time, the actual recognition system and a high correlation with that point. Confidence is a machine where the degree of confidence in recognition of the degree indicating value. Deep learning for analysis is a black box,a result of grounds and reliability of questioned the voice also many,the belief degree of the output that concerns also can reduce. Furthermore, belief degrees of the lower portion, then add in machine learning possible to improve that have been identified.
These findings, medical diagnosis and sports science, and valid indicators to provide a foothold and get. In particular, disease progression and treatment recovery, including the various elements relevant to the field and its interpretation with an aid of technology.
The findings 9 on 10 the date of the IEEE Transactions on Medical Imaging magazine online edition published in.