Development of medical aid artificial intelligence through detection of musculoskeletal X-ray abnormalities using artificial neural networks in medical image data

Main Article Content

Su Jeong Han, Choong lyeol Lee, Sung-Jun Kim

Abstract

Background / Object : Recently, the development of medical support services using artificial intelligence is in the spotlight. In this study, DICOM (Digital Imaging and Communications in Medicine) is processed and combined with patient characteristic information to perform more accurate diagnosis in disease diagnosis.


Methods / Statistical Analysis : In this study, an artificial neural network algorithm can be used to analyze musculoskeletal X-rays to quickly and simply identify patient abnormalities. Classification techniques were used to identify musculoskeletal abnormalities, and ensemble models were applied to improve the model's performance.


Findings : The abnormality in the musculoskeletal structure was classified for abnormality/normality using the DenseNet algorithm. After that, the classified data was constructed as a classification model for new data using ResNet and FusionNet algorithms. By combining the ensemble model to improve performance, we were able to classify the data with a high accuracy of 85.43%.


Improvements / Applications : Using neural network algorithms, it is possible to quickly and accurately diagnose musculoskeletal abnormalities from DICOM data, which provides powerful medical support to doctors and patients.

Article Details

Section
Articles