成果報告書詳細
管理番号20130000000199
タイトル*平成24年度中間年報 「がん超早期診断・治療機器の総合研究開発 超早期高精度診断システムの研究開発:病理画像等認識技術の研究開発 病理画像等認識自動化システムの研究開発(定量的病理診断を可能とする病理画像解析システム)」
公開日2014/6/14
報告書年度2012 - 2012
委託先名日本電気株式会社
プロジェクト番号P10003
部署名バイオテクノロジー・医療技術部
和文要約
英文要約In financial year 2012, we improved the nucleus contour extraction algorithm.
And we changed nuclei features data set for creating more sensitive hepatocellular carcinoma (HCC) detection module. As the result, the system detection was boosted the accuracy from sensitivity/specificity = 86/86% to 90/87.8%. On this system result, we investigated what kind of images becomes as false negative (miss detection). The false negative images were grouped into three patterns, first one was non-homogenous pattern images, second was high differentiated HCC (grade 1) at histological grade and last group was highly fatty tissues and/or clear cell changed patterns. As next step of improving, we created the tools for collecting the features of partial area of the images. By using partial area features, we will create the HCC detection system even on the heterogeneous pattern images. For highly differentiated HCC images, we will use the features of cord/trabecular tissue structure features which are under the development. Continuing such effort, during next financial year, we will upgrade the HCC detection accuracy level.
ダウンロード成果報告書データベース(ユーザ登録必須)から、ダウンロードしてください。

▲トップに戻る