@techreport{oai:ipsj.ixsq.nii.ac.jp:00189654,
 author = {伊東, 翼 and 太田, 佳輔 and 大石, 康博 and 村山, 正宜 and 青西, 亨 and Tsubasa, Ito and Keisuke, Ota and Yasuhiro, Oishi and Masanori, Murayama and Toru, Aonishi},
 issue = {13},
 month = {Jun},
 note = {カルシウムイメージング技術の急速な発展により，広視野から数万細胞の活動を同時記録することが可能になっている．取得される大規模イメージングデータを手動で解析するのが不可能になりつつあり，自動解析手法の需要が高まっている．特に，如何に自動細胞検出を高速かつ高精度に実現するかが課題となっている．近年，機械学習による細胞検出方法が提案されている．しかしながら，これらの手法は，近年開発されている超広視野顕微鏡で取得される大規模データを，実用的な時間で処理するのは難しい．そこで私たちは，大規模イメージングデータを実用的な時間で処理できる low-computational cost cell detection (LCCD) algorithm を開発した．実データを使い，LCCD と先行手法である Constrained Non-negative matrix factorization と Suite2P の性能を比較した．LCCD はこれら先行手法と比べて数十倍速く動作し，細胞検出性能はこれらに匹敵することを確認した．, The rapid progress of calcium imaging methods enables to record the simultaneous activity of tens of thousands of cells. However, the extreme increase in the amount of data size makes us face with a difficult of manual analysis. Consequently, a demand for automatic analysis methods for large-scale imaging data is increasing. Recently, some research groups have proposed automatic cell detection methods using machine learning approaches. Although the scalability for data size is one of the most fundamental problems in those proposed methods, within a practical time those methods cannot process large-scale data acquired with ultra-wide field-of-view microscopies recently developed. Therefore, we proposed low-computational cost cell detection (LCCD) algorithm, which can process such large-scale data within a practical time. We compared LCCD with two previously proposed methods, constrained non-negative matrix factorization and Suite2p. The detection accuracy of the LCCD was almost comparable with those of the other methods, whereas the computational time of the LCCD was a few ten times faster than those of the other methods.},
 title = {大規模なイメージングデータからの細胞検出},
 year = {2018}
}