透過您的圖書館登入
IP:3.134.104.173
  • 學位論文

磁振乳房影像之小波壓縮及其影像品質保證研究

Wavelet Compression and Image Quality Assurance on MR Breast Image

指導教授 : 林政勳
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


研究目的:本研究為?用失真小波壓縮法處?512×512×16 之乳房磁振 造影影像,觀察其可接受之壓縮比及其壓縮影像品質之保證。 材?與方法:本研究選擇100 張乳房MR 影像,包括41 張病灶在1 公分以下之影像及59 張正常無病灶之影像,MR 影像取自嘉義某醫院, 其為1.5T 之磁振造影機,此機型為專做乳房檢查而設計,每一影像大 小為 524.3 kB。先導壓縮研究先選擇10 張影像以小波壓縮軟體Apollo 分別施以10:1、15:1、20:1、25:1、30:1、40:1、50:1、55:1、60:1、65:1、 70:1、75:1、80:1 之壓縮比之影像壓縮處?,邀請有經驗之醫師判?, 選出研究時影像壓縮比?之區間。本研究將100 張影像分別做40:1、 50:1、60:1、70:1、80:1 比?之壓縮,再邀請3 位醫師做影像判?並給 予評估分?。影像品質評估吾等乃?用主觀判定法(Receiver Operating Characteristic Analysis, ROC analysis),ROC 則?用5 等第評分法,再? 用SPSS 第12 版做統計分析,並比較原影像與壓縮影像間之品質,其次 再?用Kappa 方法進?3 位醫師在判?結果一致性方面之分析。 結果:ROC 分析結果在壓縮比40:1、50:1、60:1、70:1 及80:1 之Az 值(AUC)分別為0.843、0.818、0.832、0.811 及0.833(p value<0.001) 代表40-80:1 之壓縮比?但可減少影像資?大小,亦?影響醫師判?。 ?用Kappa 方法進?3 位醫師之判?一致性分析,結果呈現A 醫師與 C 醫師在判?影像結果方面相當一致Kappa 值分別為0.467、0.540、 0.391、0.481、0.626,p value<0.001;較有差?的是B 醫師與A 醫師 在C/R50、C/R70 之判?結果Kappa 值分別為0.175、0.141,p value 分 別為0.078、0.153。在B 醫師與C 醫師之分析比較,於C/R50、C/R60 之判?結果一致性分析,Kappa 值分別為0.174、0.130,p value 分別 為0.078、0.191 一致性較有差?。 結?:當?用失真小波壓縮法在處?這?之影像時建議影像壓縮比可 為80:1,原影像大小為524.3kB,影像處?後為6.6kB 為原影像的1.25 %。

並列摘要


PURPOSE: The study was to apply wavelet compression aigorithm on 512×512×16 breast magnetic resonance (MR) images to acquire the acceptable compression ratio and their image quality. MATERIALS AND METHODS: The researcher chose 100 breast MR images from ChiaYi with 1.5 Tesla. It included 41 abnormal images with lesion size less than 1cm and 59 normal images. The original image size was 524.3kB. A pilot study prior to main study was carried out to choose the compression ratio of 40:1, 50:1, 60:1, 70:1 and 80:1 . They were ready for main study. The researcher invited three experienced radiologists to rate 1 to 5 according to the image quality on compressed images by means of Receiver Operating Characteristic Analysis ( ROC) . The intergroup agreement test was also conducted by Kappa method. RESULTS: The Area Under ROC Curve (Az) for 40:1 , 50:1 , 60:1, 70:1 and 80:1 were 0.843,0.818,0.832,0.811 and 0.833 respectively. The average Az was 0.827. It indicated such compressed images at 80:1 will retain sufficient image information to be diagnosed by radiologists. The Kappa method showed Dr.A and Dr.C with no significant differences at p value of 0.001. Yet, Dr. B was disagreement with Dr. A and Dr.C at compression ratio of 50:1, 70:1 and 50:1, 60:1 respectively. CONCLUSIONS: In the study, the accepted compressed breast MR images at a ratio of 80:1 did not adversely affect diagnostic performance or evaluation error for radiologists'' interpretation.

參考文獻


?政院衛生署(2007) http://www.doh.gov.tw.?政院衛生署國民健康局新聞
財團法人醫院評鑑暨醫?品質策進會 http://www.ticha.org.tw/96list.asp
Characteristic(ROC) analysis:Basic principles and applications in
management of newly diagnosed breast cancer.〞Arch Surg. 142(5):
Huang HK(1999)PACS-basic principle and application. Wley-Liss Press.USA

延伸閱讀