成果報告書詳細
管理番号20110000001068
タイトル*平成22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 病理画像等認識基礎技術の研究開発(定量的病理診断を可能とする病理画像認識技術)
公開日2011/8/30
報告書年度2010 - 2010
委託先名国立大学法人東京工業大学
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
和文要約和文要約等以下本編抜粋:1. 研究開発の内容及び成果等 (1) 画像特徴量を用いた数量化技術の開発 (1.1) 全スライドに対する画像特徴量数量化の基本方式 本研究開発では、デジタルスライドスキャナによって入力される全スライド画像から様々な特徴量を抽出し、マーカー発現量や臨床経過などと照合することによって診断に有益な客観的指標を求めることを目的としている。これに対して全スライドに対する画像特徴量数量化の基本方針を明確化するとともに、開発要素として組織要素の領域分割、注目領域の決定、特徴抽出、特徴量の数量化、組織分類、組織要素の認識などについて具体化した。 (1.2) 全スライドに対する特徴量の視覚化 全スライドに対する画像特徴量数量化の基礎検討として、高倍率の画像解析によって得られる画像特徴量を全スライドに対して適用し、視覚化を行った。今回は細胞核及びリンパ球の抽出に基づく特徴量を対象とした。核の抽出や核に関する特徴量の算出については既に多くの先行研究が存在するため、ここでは核抽出方法自体を研究対象とするのではなく、全スライドに対する特徴量視覚化の方法について検討することを目的とした。
英文要約Title: Research and Development Project for pathological image recognition technology, Research and Development Project to develop basic technologies for recognizing pathology images, Pathological image analysis technology to enable a quantitative pathological diagnosis, (FY2010-FY2012) FY2010 Annual Report (Tokyo Institute of Technology)
In this research, we focus on two topics; the quantification of image features and the accuracy enhancement of color and spectral information. In the topic of the quantification of image features, the techniques for automated analysis of whole slide image (WSI) were investigated. As an example of WSI analysis, the features related to nuclei were quantified such, as the density and the average size of nuclei, and the quantized features were visualized for whole slide. As a result, the difference of cancerous and non-cancerous regions was successfully visualized from the density feature of nuclei. Some other features were also quantified and show different characteristics of tissue elements. The second approach is image segmentation by a K-means clustering method based on the histogram of texture elements calculated in local areas. Analysis was performed for the images of a specimen of hepatocellular carcinoma (HCC) in magnification 5x. Experimental result showed that typical histological components, such as interstice, cancer and non-cancer areas, were approximately segmented into different classes. Another approach in the topic of image feature quantification is based on multifractal feature. Using local and global multifractal features, the cancer and non-cancer regions were classified by support vector machine. The method was experimentally applied to the images of HCC and prostate cancer. High classification rates were obtained in both HCC and prostate images. In the topic of the accuracy enhancement of color and spectral information, the color correction technique was studied using the color chart slide prepared by Massachusetts General Hospital. Since the result of preliminary experiment did not show good color correction, we should further investigate the validity of evaluation method, the characteristics of camera used here, and the sufficiency of color patches in the chart slide. In addition, for the purpose to obtain high-accuracy spectral information, an experimental system of whole slide multispectral imaging (WSMI) was developed. A software that can control the X-Y stage, Z stage, and digital camera from the connected personal computer was used, and it was confirmed that WSMI can be obtained by tiling partial images. As the limitation of the software used here, the size of merged image was limited, thus the development of new software for tiling larger images is one of the issues of next step.
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