Department of Radiology, Seoul St. Mary's Hospital, College of Medicine, The Catholic University of Korea, 505 Banpo-Dong, Seocho-Ku, Seoul 137-040, Republic of Korea.
Clin Radiol. 2010 Dec;65(12):967-73. doi: 10.1016/j.crad.2010.02.001. Epub 2010 Sep 15.
To determine whether radiologists can recognize images retouched to include sham lesions.
Ten representative key images were selected of aortic dissection, hepatocellular carcinoma, renal cell carcinoma, colon cancer, liver metastasis, hepatic cyst, gallbladder stones, splenic artery aneurysm, adrenal adenoma, and stomach cancer from abdominal computed tomography (CT) imaging performed in 2008. Five of the key images were replaced with retouched images using image-editing software. The time to complete retouching was recorded for each image. Radiologists were requested to make a diagnosis for the 10 images, and were then asked to identify possible retouched images. The time taken to reach a decision in each case was recorded. Thirty radiologists (13 residents and 17 attending radiologists) participated as reviewers.
The time to complete retouching was 15.2±3.15 min. None of the reviewers recognized that some images were retouched during diagnosis. The rate of correct diagnosis was 90% (range 71.7-100%). After reviewers were informed of possible image retouching, the detection rate of retouched images was 50% (40-58.3%). This rate was statistically the same as random choice (p=0.876). There was no significant difference between residents and attending radiologists in the detection rate of retouched images (p=0.786). The time to diagnosis and the time to detection of the retouched images were 15 (14-17) and 6 (5-7) min, respectively.
Digital images can be easily retouched, and radiologists have difficulty in identifying retouched images. Radiologists should be aware of the potential fraudulent use of retouched images.
确定放射科医生是否能够识别经过修饰以包含虚假病变的图像。
从 2008 年进行的腹部计算机断层扫描(CT)成像中选择了 10 个代表性的关键图像,包括主动脉夹层、肝细胞癌、肾细胞癌、结肠癌、肝转移、肝囊肿、胆囊结石、脾动脉瘤、肾上腺腺瘤和胃癌。使用图像处理软件将其中 5 个关键图像替换为修饰后的图像。记录每张图像的修饰完成时间。要求放射科医生对这 10 张图像进行诊断,然后要求他们识别可能经过修饰的图像。记录每个病例做出决策所花费的时间。30 名放射科医生(13 名住院医师和 17 名主治放射科医生)作为评审员参与。
完成修饰的时间为 15.2±3.15 分钟。在诊断过程中,没有评审员意识到有些图像经过了修饰。正确诊断率为 90%(范围为 71.7-100%)。在告知评审员可能存在图像修饰后,对修饰图像的检测率为 50%(40-58.3%)。这一比率与随机选择的比率相同(p=0.876)。在检测修饰图像的能力方面,住院医师和主治放射科医生之间没有统计学差异(p=0.786)。诊断时间和检测修饰图像的时间分别为 15(14-17)和 6(5-7)分钟。
数字图像可以很容易地进行修饰,放射科医生很难识别修饰后的图像。放射科医生应该意识到修饰图像可能被用于欺诈。