本論文目的在於研究經過濾波器處理後MRI影像之影像品質差異及理想的測量方法。 本研究選擇以80張頭部MRI影像,其中包括T1、T2、T1 FLAIR和 T2 FLAIR影像序列等每種各20張。以系統中提供的影像濾波器,分別將原始影像處理後得到較佳的影像再以客觀方法判定。本研究使用了六種的客觀影像品質測量方法來對醫學影像模糊程度做分析,分別有以點對點為基礎(pixel based-metric,PBM)的有均方誤差(Mean Square Error,MSE)、雜訊比SNR(Signal to Noise Ratio,SNR)、峰值雜訊比(Peak Signal to Noise Ratio,PSNR)等三種的方法,另外以視窗作為計算基礎(Window Based-Metric,WBM)的方法包括UQI(Universal Quality Index)、MSSIM(Mean Structural Similarity)與MPR(Moran’s Peak Ratio)等三種方法。 結果發現以點為基礎的測量法中雖然對於影像品質的降級很敏感,但卻無法提供其相關資訊,藉由局部區域進行的以視窗作為計算基礎的方法可以提供影像變異的訊息,尤其是MPR的方法。透過濾波器處理的各個序列,可由MPR的結果得知T1和T2影像的差異。 由於MPR對於MRI的濾波過的影像品質相當敏感,因此本研究建議採MPR做為未來測量MRI影像經濾波器處理的影像品質評估方法。
The purpose of this study was to seek an optimal method for measuring the image quality of MRI images after filtering. The study chose 80 slices brain image of MRI. They included T1WI, T2WI, T1 Flair and T2 Flair series. The original images were processed with the filter of the system, and to analyze them with objective method. In this work, the authors applied different level of filtering for brain MRI images. There were six objective methods to evaluate image quality, one was Pixel Based-Metric (PBM) and the other was Window Based-Metric (WBM). The PBM included Mean Square Error (MSE), Signal to Noise Ratio (SNR), Peak Signal to Noise Ratio (PSNR) meanwhile Universal Quality Index (UQI), Mean Structural Similarity (MSSIM), and Moran’s Peak Ratio (MPR) was for WBM. The PBM was very sensitive to image degradation but didn’t correlate well to subjective quality measures. The image spatial information was estimated from a local region by window based-metric, especially in MPR measurement. Apply the filter process by series, the researcher found that the image showed differences of MPR between T1and T2. The MPR was shown to be very sensitive to image quality of MRI filter, MPR was recommend as for measuring the image quality of MRI images after filtering.