機械学習を用いた骨強度評価モデルの検討

抄録

In Japan, it is estimated that there are about 12.8 million people who have osteoporosis, and it is expected that the number of osteoporosis patients may continue to increase. DXA is used to measure bone strength, however bone strength using DXA may not be measured accurately, and it is difficult to measure it in terms of safety and ease. In this study, we have performed machine learnings to predict bone strength obtained by the hammering test of bones. To predict bone strength, age, body composition, and physical fitness test data were used as explanatory variables. The accuracy of models using linear regression, decision tree and neural network were compared. It was found that the model using a decision tree showed the highest accuracy. It was confirmed that the same prediction values were obtained in the case of the decision tree model, though the measurement values were different. Further modeling is still necessary to optimize the accuracy of the predictions.