University Hospital Zurich, Institute of Diagnostic and Interventional Radiology, Rämistrasse 100 8091 Zürich Switzerland.
University Hospital Zurich, Institute of Diagnostic and Interventional Radiology, Rämistrasse 100 8091 Zürich Switzerland.
Eur J Radiol. 2023 Apr;161:110733. doi: 10.1016/j.ejrad.2023.110733. Epub 2023 Feb 9.
To evaluate different image quality scoring systems in the assessment of factors limiting diagnostic accuracy of prostate MRI.
This retrospective IRB-approved study included 281 patients undergoing prostate MRI prior to biopsy. Four readers (2 experienced, 2 novice) independently reviewed all MRI examinations (n = 295) and assigned scores for subjective image quality (1-5; 1:poor, 5:excellent), the PI-QUAL and the PSHS scoring system. The original PI-RADS scores were extracted from the report and transperineal template saturation biopsy served as histopathological reference.
Inter-reader agreement was found to be good, with PSHS showing highest agreement (kappa: 0.65). The PSHS scoring system performed well assessing the influence of image quality on sensitivity of MR for clinically-significant cancer for the experienced readers using a PI-RADS score cut-off ≥ 3/≥4, as did the PI-QUAL scoring system with a PI-RADS cut-off ≥ 4. For the less experienced radiologist, this was true for PSHS (clinically-significant and all cancers) and PI-QUAL scores (clinically-significant cancers) for a PI-RADS score ≥ 3. PSHS scores were positively associated with the detection of clinically-significant cancer based on a PI-RADS cut-off ≥ 4, OR 1.86 (95 % CI 1.22-2.82), and had the highest Somers' D.
The PSHS scoring system performed well in assessing the effect of image quality on detection rates, as did the PI-QUAL system. Since both systems focus on different aspects of image quality, their incorporation into prostate MRI reports could further enhance standardization and allow for a reliable assessment of image quality as a potential confounder in prostate MRI.
评估不同的图像质量评分系统在评估前列腺 MRI 诊断准确性限制因素中的作用。
本回顾性 IRB 批准的研究纳入了 281 例在活检前接受前列腺 MRI 检查的患者。4 名读者（2 名有经验，2 名无经验）独立评估了所有 MRI 检查（n=295），并对主观图像质量（1-5；1：差，5：优）、PI-QUAL 和 PSHS 评分系统进行评分。原始 PI-RADS 评分从报告中提取，经会阴模板饱和活检作为组织病理学参考。
发现读者间的一致性较好，PSHS 显示出最高的一致性（kappa：0.65）。在使用 PI-RADS 评分≥3/≥4 的界值时，PSHS 评分系统在评估图像质量对有临床意义的癌症的 MRI 敏感性的影响方面表现良好，PI-QUAL 评分系统在使用 PI-RADS 评分≥4 的界值时也表现良好。对于经验较少的放射科医生，当 PI-RADS 评分≥3 时，PSHS（有临床意义和所有癌症）和 PI-QUAL（有临床意义的癌症）评分也是如此。PSHS 评分与基于 PI-RADS 界值≥4 的有临床意义的癌症检出率呈正相关，OR 为 1.86（95%CI 1.22-2.82），且具有最高的 Somers' D。
PSHS 评分系统在评估图像质量对检出率的影响方面表现良好，PI-QUAL 系统也是如此。由于这两个系统都关注图像质量的不同方面，将它们纳入前列腺 MRI 报告中可以进一步增强标准化，并允许可靠地评估图像质量作为前列腺 MRI 的潜在混杂因素。
