Article

目的:本研究では,濃度ヒストグラム解析法を用いて,急性期脳梗塞の脳MRI拡散強調画像(DWI)の表示条件を自動設定する方法を考案した.方法:本研究では,虚血後6時間以内に撮像された60症例のDWIを利用した.本手法は,まず,別の1症例に対し,b0画像上の視床の信号強度を利用してDWIの表示条件を設定させる既存方法を適用し,線形濃度変換法を用いて表示条件を8bitsに変換した.次に,濃度ヒストグラムを求めて,最大頻度値の半値となる画素値の全幅(FWHM)を求めた.その後,対象の60症例に対して3次元DWIを作成し,脳実質部における濃度ヒストグラムを求めて,最大頻度値となる画素値を決定した.最後に,決定された画素値をウィンドウ中心として設定し,調節されたDWIの濃度ヒストグラム上のFWHMと先に求めたFWHMとが同一となるウィンドウ幅を求めて設定し,3次元DWIの表示階調を調節した.既存の手動法と本自動化法を比較するため,手法ごとで準備した画像のペアに対する2肢強制選択法(2-AFC法)を実施し,信号強度と画像コントラストの変動を主観的に評価した.結果:2-AFC法を実施し,本手法により調節された3次元DWIの症例間に対する主観的類似度を求めた結果,選択率が71.7%となり,本自動化法は,既存方法に比べ,より一貫した画像表示を提供できることが明確に示された.結論:本手法は,脳DWIの表示条件の正確且つ迅速な自動設定において有望な技術であり,画像診断における放射線科医の意思決定にとって有用である.

Objectives : In this study, we evaluated a computerized scheme for automated setting of grayscale display for diffusion-weighted magnetic resonance images (DWI) of brains based on the density histogram analysis against existing standard methods for setting display conditions for DWI to ascertain whether the automated methods provided more consistent display. Methods : We used 60 brain DWI cases, which were scanned within the first 6 hours after stroke onset. In our computerized scheme, the existing method of setting the display conditions for DWI using the thalamic signal intensity in a b0 image was first applied for one DWI case. The display grayscale was converted into 8-bits by the linear grayscale transfer technique. The density histogram was then determined, and the full width of pixel values corresponding to half of the maximum (FWHM) in density histogram was determined. The DWI volumes were then created for 60 cases, and the density histogram was determined from the brain region based on DWI volumes. The pixel value corresponding to the maximum in density histogram was determined for DWI volume. The display grayscale for DWI volumes was adjusted by setting the determined pixel value as a window center, and the pixel value that FWHM in the density histogram of adjusted DWI is corresponding to FWHM previously obtained as a window width. To compare the existing manual method to the automated method, the variation in signal intensity and image contrast was evaluated subjectively using a two-alternative-forced-choice test for pairs of images prepared by each method. Results : The subjective similarity for pairs of DWI volumes adjusted using our computerized scheme was judged to be more similar than those using existing manual methods by 71.7%, which indicated clearly our automated method provided more consistent display. Conclusions : Our computerized scheme is a promising technique for accurate, prompt automated setting of display conditions in brain DWI, and may be useful to radiologists in decision-making for radiological diagnosis.

原著

群馬県立県民健康科学大学紀要 6,33-40,2011