@techreport{oai:ipsj.ixsq.nii.ac.jp:00211731,
 author = {佐藤, 孝治 and 林, 昌純 and 清元, 佑紀 and 重田, 恵吾 and 松本, 省二 and 小山, 裕司 and Takaharu, Sato and Masazumi, Hayashi and Yuki, Kiyomoto and Keigo, Shigeta and Shoji, Matsumoto and Hiroshi, Koyama},
 issue = {28},
 month = {Jun},
 note = {本研究は機械学習技術を用いて急性脳主幹動脈閉塞症の予測精度を向上，ならびにアプリケーションの操作性を改善する試みである．先の研究で我々は機械学習による ELVO 予測アプリケーションを開発し，感度 75%，偽陰性率 25% であった．今回の実装では，1．ELVO 症例数の追加（86 症例），2．拡張期血圧を連続値から 10 刻みへ変更，3．年齢を連続値から離散値へ変更，4．決定木の深さを調整，5．クラスの重みを調整，6．決定木の数の変更を試行し，それぞれ比較した．その結果，1，3，5が有効であることを特定し，感度 93.1%，偽陰性率 6.9% を実現した．また，2 の変更は感度と偽陰性率の向上には寄与しないがアプリケーションの操作性を改善できるため採用した．, This study is an attempt to improve the prediction level of Emergent Large Vessel Occlusion (ELVO) by machine learning methods and the usability of the application, called ELVO checker. In our previous result, we developed a machine learning-based ELVO prediction application with a sensitivity of 75% and a false negative rate of 25%. In this paper, we attempted (1) to add ELVO 86 cases, (2) to change the diastolic blood pressure from continuous to 10 increments, (3) to change the age from continuous to discrete, (4) to adjust the maximum depth of decision trees (5) to adjust the class weights, and (6) to change the number of decision trees, and compared these results. As a result, we found the change 1, 3, and 5 were effective and obtained a sensitivity of 93.1% and a false negative rate of 6.9%. The change 2 was not effective, but was adopted since it was expected to improves the usability of the application.},
 title = {機械学習による急性脳主幹動脈閉塞症の予測精度の向上とアプリケーションの操作性改善},
 year = {2021}
}