Computer-aided diagnosis for prostate cancer detection in multiparametric MRI:a multireader multicase study
10.3969/j.issn.1002-1671.2024.05.034
- VernacularTitle:计算机辅助诊断对前列腺癌多参数MRI诊断的影响:多读片者多病例研究
- Author:
Ge GAO
1
;
Kexin WANG
;
Huihui WANG
;
Rui WANG
;
Yi LIU
;
Xiaoying WANG
Author Information
1. 北京大学第一医院医学影像科,北京 100034
- Keywords:
computer-aided diagnosis;
prostate cancer;
mul-tiparametric magnetic resonance imaging
- From:
Journal of Practical Radiology
2024;40(5):825-829
- CountryChina
- Language:Chinese
-
Abstract:
Objective To investigate the effect of computer-aided diagnosis(CAD)as a"second reader"integrated in prostate cancer(PCa)multiparametric magnetic resonance imaging(mpMRI)on the diagnostic performance.Methods Sixty-four patients were recruited retrospectively.Six readers were divided into 3 groups according to the experience.Unknown the pathologic diagnosis,readers graded according to prostate imaging reporting and data system(PI-RADS)v2.1 and accepted or rejected the revision of PI-RADS v2.1 scores by CAD.Using pathology as the gold standard,the diagnostic efficiency,diagnostic confidence and reading time of PI-RADS v2.1 were compared before and after CAD assistant.Results On the level of lesions,without CAD,the area under the curve(AUC)of reader A1 was higher than B1,C1 and C2,and the AUC of reader A2 was higher than C1 and C2,the difference was statistically significant(P<0.05).While the difference of AUC among other readers were not significant(P>0.05).With CAD,the difference of AUC among all readers were not significant(P>0.05).On the level of patients,the mean AUC of all readers with CAD[0.903,95%confidence interval(CI)0.848-0.958]was higher than that without CAD(0.863,95%CI 0.800-0.929),the differ-ence was statistically significant(F=6.52,P=0.040).The difference of AUC between without and with CAD assistance was-0.040(95%CI-0.078--0.003),standard deviation was 0.016,the difference was statistically significant(t=-2.55,P=0.040).The diag-nostic confidence of readers was improved significantly with CAD(P<0.05).Besides,the average reading time of each case required an additional 0.8 min.Conclusion After integrating CAD into PCa mpMRI diagnosis as a"second reader",the performance of less experienced readers can be improved and show similar diagnostic efficiency with experienced readers.