{"created":"2023-05-15T14:48:39.594596+00:00","id":66662,"links":{},"metadata":{"_buckets":{"deposit":"fbd4b408-c4a6-4f16-8111-e472263029fb"},"_deposit":{"created_by":1,"id":"66662","owners":[1],"pid":{"revision_id":0,"type":"depid","value":"66662"},"status":"published"},"_oai":{"id":"oai:repo.qst.go.jp:00066662","sets":["10:29"]},"author_link":["655590","655595","655592","655588","655596","655593","655585","655594","655586","655589","655591","655587"],"item_10005_date_7":{"attribute_name":"発表年月日","attribute_value_mlt":[{"subitem_date_issued_datetime":"2017-10-16","subitem_date_issued_type":"Issued"}]},"item_10005_description_5":{"attribute_name":"抄録","attribute_value_mlt":[{"subitem_description":"[背景]\nDeep Learning (convolutional Deep Neural Network: cDNN)は画像分類などの機械学習を高精度・半自動的に行えるが, 判断根拠が不透明な問題がある. cDNNによる臨床画像分類で, 精度を確保しつつ判断根拠を可視化する方法を検討した.\n[方法]\n健常脳MRI画像をcDNNにより2つの年齢群に分類する実験を行った. 対象画像(500例)から異なる解剖学的領域が抽出された5つの画像群を作成した(図中A-E). それぞれを学習データと試験データに分け, 得られた5種の学習データを個別のcDNNに与えて画像を2つの年齢群へ分類する学習をさせた(並列学習). また, これらのcDNNの最終層出力を結合しlogistic回帰による分類学習を行った (統合学習, 図中F). 試験データを正しく分類する率を指標とし各学習の正確さを統計学的に比較した(Wilcoxon符号順位検定, P<.05を有意).<br>\n[結果]\n並列学習では抽出領域の違いにより正答率に差を認めた. この結果から, どの領域が学習に重要であったかが逆説的に示唆される(例: BとCの有意差から実質外腔が重要). また, 統合学習により正答率は向上する傾向を認めた(BとFの比較でP= .06). 並列・統合学習の併用は判断根拠の可視化と精度確保の両立につながる.","subitem_description_type":"Abstract"}]},"item_10005_description_6":{"attribute_name":"会議概要（会議名, 開催地, 会期, 主催者等）","attribute_value_mlt":[{"subitem_description":"第45回日本磁気共鳴医学会大会","subitem_description_type":"Other"}]},"item_access_right":{"attribute_name":"アクセス権","attribute_value_mlt":[{"subitem_access_right":"metadata only access","subitem_access_right_uri":"http://purl.org/coar/access_right/c_14cb"}]},"item_creator":{"attribute_name":"著者","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"立花, 泰彦"}],"nameIdentifiers":[{"nameIdentifier":"655585","nameIdentifierScheme":"WEKO"}]},{"creatorNames":[{"creatorName":"小畠, 隆行"}],"nameIdentifiers":[{"nameIdentifier":"655586","nameIdentifierScheme":"WEKO"}]},{"creatorNames":[{"creatorName":"漆畑, 拓弥"}],"nameIdentifiers":[{"nameIdentifier":"655587","nameIdentifierScheme":"WEKO"}]},{"creatorNames":[{"creatorName":"尾松, 徳彦"}],"nameIdentifiers":[{"nameIdentifier":"655588","nameIdentifierScheme":"WEKO"}]},{"creatorNames":[{"creatorName":"岸本, 理和"}],"nameIdentifiers":[{"nameIdentifier":"655589","nameIdentifierScheme":"WEKO"}]},{"creatorNames":[{"creatorName":"東, 達也"}],"nameIdentifiers":[{"nameIdentifier":"655590","nameIdentifierScheme":"WEKO"}]},{"creatorNames":[{"creatorName":"立花 泰彦","creatorNameLang":"en"}],"nameIdentifiers":[{"nameIdentifier":"655591","nameIdentifierScheme":"WEKO"}]},{"creatorNames":[{"creatorName":"小畠 隆行","creatorNameLang":"en"}],"nameIdentifiers":[{"nameIdentifier":"655592","nameIdentifierScheme":"WEKO"}]},{"creatorNames":[{"creatorName":"漆畑 拓弥","creatorNameLang":"en"}],"nameIdentifiers":[{"nameIdentifier":"655593","nameIdentifierScheme":"WEKO"}]},{"creatorNames":[{"creatorName":"尾松 徳彦","creatorNameLang":"en"}],"nameIdentifiers":[{"nameIdentifier":"655594","nameIdentifierScheme":"WEKO"}]},{"creatorNames":[{"creatorName":"岸本 理和","creatorNameLang":"en"}],"nameIdentifiers":[{"nameIdentifier":"655595","nameIdentifierScheme":"WEKO"}]},{"creatorNames":[{"creatorName":"東 達也","creatorNameLang":"en"}],"nameIdentifiers":[{"nameIdentifier":"655596","nameIdentifierScheme":"WEKO"}]}]},"item_language":{"attribute_name":"言語","attribute_value_mlt":[{"subitem_language":"jpn"}]},"item_resource_type":{"attribute_name":"資源タイプ","attribute_value_mlt":[{"resourcetype":"conference object","resourceuri":"http://purl.org/coar/resource_type/c_c94f"}]},"item_title":"Deep learning を利用した画像分類において学習精度と判断過程の可視化を両立する手法の検討","item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"Deep learning を利用した画像分類において学習精度と判断過程の可視化を両立する手法の検討"}]},"item_type_id":"10005","owner":"1","path":["29"],"pubdate":{"attribute_name":"公開日","attribute_value":"2018-03-19"},"publish_date":"2018-03-19","publish_status":"0","recid":"66662","relation_version_is_last":true,"title":["Deep learning を利用した画像分類において学習精度と判断過程の可視化を両立する手法の検討"],"weko_creator_id":"1","weko_shared_id":-1},"updated":"2023-05-15T20:46:24.670608+00:00"}