Accuracy, intra-, and inter-radiologist variability of PI-RADS v2.1 scoring for clinically significant prostate cancer detection

Introduction

According to the 2020 global cancer (GLOBOCAN) data (1), there were almost 1.4 million new cases of prostate cancer (PCa) worldwide. It is the most frequently diagnosed cancer in men in 112 countries/regions, and the fifth leading cause of male tumor mortality. The invasiveness and harm associated with prostate lesions, including benign prostatic hyperplasia (BPH) and PCa, are highly variable. The aggressive subtype of PCa (Gleason score 7–10) is defined as clinically significant PCa and prone to metastatic progression and even death, requiring timely treatment and active intervention (2,3). Conversely, the less aggressive subtype consists of clinically non-significant PCa and BPH. Active surveillance, rather than surgery, should be implemented for disease management (4-6). Consequently, accurate identification of clinically significant lesions is critical for selecting treatment options for prostate diseases.

Magnetic resonance imaging (MRI) provides multiplanar imaging and a wide selection of sequences. It has become one of the most important evaluation methods for the preoperative diagnosis and evaluation of the prostate (7). The prostate imaging reporting and data system (PI-RADS) has been proposed to ensure the consistency of the diagnosis, management, and communication of PCa. This system outlines the observation sequence and related scoring method of suspicious prostate lesions at different locations to obtain the PI-RADS category. The PI-RADS category ranges between 1 and 5, indicating a very low and a very high likelihood that a lesion is clinically significant PCa, respectively. It is constantly updated based on existing evidence, with the European Association of Urology releasing multiple versions, including version 1.0 in 2011, version 2.0 in 2015, and version 2.1 in 2019 (8-10).

Some researchers have reported that different readers differ in the PI-RADS category of the same lesion due to the discrepancy in educational background, imaging reading habits, and subjective definitions. It has been reported that the inter-agreement of PI-RADS v2.0 category between different readers is only moderate, both in the transition zone or the peripheral zone (11,12). Another study that used the PI-RADS v2.1 category to identify significant PCa reported that the inter-observer agreement of the PI-RADS category obtained by radiologists with different levels of experience was moderate to strong (13). Differences in the PI-RADS category among different radiologists can lead to differences in management modalities for the same patient, such as whether to perform a puncture or not, thereby causing confusion for urologists. Determining the source of score heterogeneity would help to optimize and improve the subsequent PI-RADS category, thus assisting more stable treatment decision-making. The above studies only measured the agreement of categories by radiologists from the same institution. The degree of agreement of categories, as well as the sequence score, among radiologists from different institutions is still unknown. The purpose of this study was to explore the accuracy of radiologists with various levels of experience from different institutions in identifying prostate lesions via the PI-RADS v2.1 and to further determine the inter- and intra-observer agreement of sequence scores and categories. We present this article in accordance with the STARD reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-37/rc).

Methods

Case collection

This study retrospectively analyzed consecutive patients who underwent prostate MRI at The Second Affiliated Hospital, Jiangxi Medical College, Nanchang University between 1 July 2022, and 30 June 2023 due to suspected prostate disease. The inclusion criteria were as follows: (I) patients who underwent preoperative prostate magnetic resonance (MR) examination; and (II) patients whose pathology results were obtained via needle biopsy or prostatectomy within one month of MRI scanning, with a Gleason score determined for PCa.

The exclusion criteria were as follows: (I) patients who had received prostate-related treatments (including endocrine, chemoradiotherapy, or surgical treatment) before MRI; (II) patients with incomplete imaging data, such as a lack of T2-weighted imaging (T2WI) or diffusion-weighted imaging (DWI) sequences; and (III) patients with poor image quality, which would affect the assessment of the lesion.

PCa patients with a Gleason score ≥7 according to the pathological results were included in the clinically significant group. BPH and PCa patients with a Gleason score <7 were included in the non-significant group. A total of 164 patients, including 65 in the significant subgroup and 99 in the non-significant group, were enrolled in the final analysis. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by The Second Affiliated Hospital of Nanchang University Medical Research Ethics Committee (No. IIT-0-2024-279) and the requirement for individual consent for this analysis was waived due to the retrospective nature.

Image acquisition

All prostate MR images were acquired on 3 MRI machines (1.5 T, GE Signa HDxt; 3.0 T MRI, GE Signa HDxt; 3.0 T MRI, GE Discovery; GE Healthcare, Chicago, IL, USA). Prostate MRI included T1-weighted imaging (T1WI), T2-weighted imaging (T2WI), and DWI sequence imaging, and some of the patients underwent dynamic contrast enhancement (DCE) imaging. The b values of DWI sequence acquisition were 0, 1,000, 2,000, and 3,000 s/mm². The field of view (FOV) was 260 mm × 260 mm, and the specific parameters are shown in Table S1. The DCE sequence was acquired via intravenous injection of gadopentetate dimeglumine at an injection rate of 2.5 mL/second (0.1 mmol/kg). After the contrast agent was injected, 30 consecutive dynamic scans were performed.

Image evaluation

The imaging assessment was conducted independently by six radiologists from 5 hospitals. Readers 1, 2, and 3 were radiologists with 21, 8, and 7 years of work experience, respectively (senior group); readers 4, 5, and 6 were radiologists with 4, 3, and 3 years of working experience (junior group), respectively. All the readers were trained by an experienced radiologist (28 years of experience) according to the guidelines of the “Prostate Imaging Reporting and Data System Version 2.1: 2019 Update of Prostate Imaging Reporting and Data System Version 2” before image interpretation (8).

All the images were exported in Digital Imaging and Communications in Medicine (DICOM) format and then imported into the RadiAnt DICOM Viewer (Medixant, RadiAnt DICOM Viewer Version 2021.2). Each reader independently scored all the images when they were blinded to the clinical information. The scoring steps were in strict accordance with PI-RADS v2.1 criteria. Only the scores of the sequences (T2WI, DWI/ADC, and DCE) obtained during the process and the final PI-RADS categories were recorded. For the patients (glands) with 2 or more suspected lesions, the radiologists were asked to record only the scores of the lesion that he/she thought would have the highest probability of malignancy. One month later, all the readers reevaluated the images again referring to the design of the previous study (14). Figure 1 shows 2 typical cases.

Figure 1 Two representative examples evaluated by PI-RADS v2.1. In patient 1 (A-D), a suspicious lesion was found in the transition zone (red arrows). The T2WI score (A) was evaluated according to PI-RADS v2.1. The scores of the 6 readers were 3, 3, 2, 2, 2, and 3, respectively. The DWI score was subsequently obtained by observing the DWI (B) and the ADC map (C). The scores judged by the 6 readers were 3, 4, 3, 3, 4, and 4, respectively. There was no need to analyze the DCE (D). The final PI-RADS categories from the readers were 3, 3, 2, 2, 3, and 3, respectively. The pathological results suggested that the lesion was benign prostatic hyperplasia, which belongs to the non-significant group. In patient 2 (E-H), a suspicious lesion was found in the peripheral zone (green arrows). According to PI-RADS v2.1, the DWI score was evaluated by observing the DWI (F) and ADC maps (G). The scores of the 6 readers were 5, 5, 4, 5, 4, and 4, respectively. There was no need to score the T2WI (E) or the DCE (H). The final PI-RADS categories of the readers were 5, 5, 4, 5, 4, and 4, respectively. The pathological results suggested that the lesion was prostate cancer with a Gleason score of 8, which belongs to the significant group. ADC, apparent diffusion coefficient; DCE, dynamic contrast enhancement; DWI, diffusion-weighted imaging; PI-RADS, Prostate Imaging Reporting and Data System; T2WI, T2-weighted imaging.

Statistical analysis

All the statistics and graphics were generated via the software SPSS 29.0.1.0 (IBM Corp., Armonk, NY, USA) and GraphPad Prism version 10.2.3 (GraphPad Software, San Diego, CA, USA). The Kolmogorov-Smirnov test was used to test for a normal distribution. The results were expressed as the means ± standard deviations if the data were normally distributed, and the differences among groups were compared via independent samples t-test. The values that did not conform to a normal distribution were expressed as the median and interquartile range (IQR). The differences among groups were compared via the Mann-Whitney U test. Receiver operating characteristic (ROC) curves were used to evaluate the performance of the PI-RADS categories in distinguishing between significant and non-significant lesions. The area under the curve (AUC), sensitivity, and specificity were obtained.

The weighted kappa coefficient was used to evaluate the intra-reproducibility of the DWI score, T2WI score, and PI-RADS category by the same reader between the 2 rounds of evaluation. The Kendall W coordination coefficient (KW value) was used to evaluate the inter-observer consistency of the sequence score and category. The weighted kappa coefficient and the Kendall W coefficient were interpreted as follows: 0–0.20, slight agreement; 0.21–0.40, fair agreement; 0.41–0.60, moderate agreement; 0.61–0.80, substantial agreement; and 0.81–1.00, almost perfect agreement (15,16).

Results

General information

A total of 175 patients met the inclusion criteria. Among them, 6 patients were excluded because of previous endocrine therapy, 2 were excluded because of a history of electric cutting surgery, and 3 were excluded because of incomplete imaging data (Figure 2). Thus, 164 patients were enrolled. The mean age was 70.2±7.9 years (range 52–91 years). The significant group consisted of 65 patients, with a mean age of 70.20±7.25 years and a median prostate-specific antigen (PSA) level of 12.00 (IQR 7.30, 27.11) ng/mL. The Gleason score was 7 for 28 patients, 8 for 26 patients, 9 for 10 patients, 10 for 10 patients, and 11 for 1 patient. The non-significant subgroup included 99 patients; the mean age was 70.21±8.40 years, and the median PSA level was 12.80 (IQR 8.13, 27.15). There were 12 patients with non-significant PCa and 87 with BPH. No difference was found in age (P=0.092) or PSA level (P=0.632) between the 2 groups.

Figure 2 Flow of participants selection. MRI, magnetic resonance imaging.

Intra-observer agreement

The weighted kappa coefficients between the 2 rounds of scoring of the 6 radiologists are shown in Table 1. The T2WI score, DWI score, and PI-RADS category had moderate to substantial agreement (K values, 0.408–0.763; all P<0.001). The intra-observer consistency of the senior group reached substantial agreement (kappa coefficient, 0.612–0.763), whereas the junior group showed moderate intra-observer consistency agreement (weighted kappa coefficient, 0.408–0.685).

Inter-observer agreement

The KW values of the T2WI, DWI, and PI-RADS scores obtained by all the radiologists were 0.199, 0.106, and 0.127, respectively (all P<0.001). In the senior group, the KW values for the scores and categories were 0.279, 0.103, and 0.147, respectively (all P<0.001). In the junior group, the KW values were 0.143, 0.089, and 0.139, respectively (all P<0.001). Except for the consistency of the T2WI score in the senior group, which reached fair agreement, the other scores showed poor agreement.

Accuracy of PI-RADS in identifying significant lesions

The detailed PI-RADS categories determined by the 6 radiologists and the number of clinically significant lesions confirmed by pathology in each category are shown in Table 2. All the prostate glands rated as category 1 were non-significant. The percentages of significant PCa cases in categories 2 and 3 were 2.9–16.2% and 8.1–38.7%, respectively. Non-significant lesions accounted for 18.2–45.6% and 8.5–22.2% of category 4 and 5 lesions, respectively. Significant lesions tend to be assigned to higher categories. Non-significant diseases made up over half of category 3, and the proportion of non-significant lesions in category 4 was still high.

The AUCs of the PI-RADS categories identified by the six radiologists for identifying the significance of prostates were 0.915, 0.881, 0.863, 0.849, 0.833, and 0.884, respectively (P<0.001). The corresponding parameters, such as sensitivity and specificity, are shown in Table 3. The ROC curves are shown in Figure 3. Among them, Reader 1 obtained the highest AUC and sensitivity.

Figure 3 ROC curves were used to evaluate the efficacy of 6 radiologists in identifying the clinical significance prostate lesions by PI-RADS category. PI-RADS, Prostate Imaging Reporting and Data System; ROC, receiver operating characteristic.

Discussion

This study evaluated the accuracy of PI-RADS v 2.1 in identifying significant prostate lesions, as well as the intra- and inter-repeatability of scoring by 6 radiologists. The evaluation was conducted by 6 readers from 5 hospitals, including 3 experienced senior radiologists and 3 inexperienced junior radiologists. The results revealed that both senior and junior radiologists performed well in the identification of significant lesions using the PI-RADS v2.1 category, with senior readers performing better than junior readers (AUC, 0.863–0.915 vs. 0.833–0.884). The PI-RADS and sequence scores of the same reader showed moderate to substantial agreement, and the reproducibility of the senior group was greater than that of the junior group. Inter-observer agreement for the T2WI score in the senior group was fair; however, agreement for the T2WI score in the junior group, DWI scores, and PI-RADS categories overall and within the subgroups was poor.

Thompson et al. reported that the negative predictive value of the PI-RADS v1.0 category in identifying significant PCa lesions was as high as 93–96%, but the positive predictive value was low (47–53%) (17). Another study further determined the threshold of the PI-RADS category for defining significant and non-significant lesions, and the results revealed that when the category was ≤2, it had a very high negative predictive value (97.7%) and high accuracy (18). Our results were consistent with their results. The AUCs of the PI-RADS category for the identification of significant lesions reached 0.833, and that of some radiologists reached 0.9 or above, suggesting that PI-RADS v2.1 category has strong practicability. Different radiologists excelled in different aspects of lesion evaluation; some had extremely high sensitivity (reader 1, 92.3%), and some had extremely high specificity (reader 5, 92.4%).

The European Association of Urology guidelines recommend prostate biopsy for patients with a PI-RADS score ≥3 (19). Our study further revealed that all the lesions in PI-RADS category 1 were non-significant lesions. Most of the cases were correctly diagnosed in categories 2 (non-significant lesions, 93.8–97.1%) and 5 (significant lesions, 77.7–91.5%). However, the proportions of significant lesions in category 3 and non-significant lesions in category 4 were 8.1–38.7% and 8.2–45.6%, respectively, which was a wide variety and relatively high. These results are consistent with those of Wei et al., who reported similar proportions of non-significant lesions with categories of 3 and 4–5 (20). Such a classification would result in a large number of non-significant lesions, especially in categories 3 and 4, undergoing unnecessary invasive puncture. This phenomenon has also been a concern of many clinical scientists. Some studies have proposed that lesions with a category of 3 should not be punctured and that MRI follow-up should be used instead (21). Certain studies have aimed to enhance the precision of assessing this patient group by integrating factors such as ADC value and age (22).

This study also investigated the consistency and repeatability within and among readers. Unlike previous studies, we analyzed not only the agreement of the PI-RADS category but also the sequence (DWI and T2WI) scores to determine the factors that cause variation to promote the optimization of the PI-RADS. Given that fewer cases with DCE scores were evaluated by readers at the same time, we did not conduct consistency analysis for this sequence because the reliability of the results could be questioned. In terms of intra-observer stability, the scores of the 2 consecutive observations by the same radiologist had moderate to substantial agreement. The reproducibility of the senior group reached a high level in both the sequence scores and categories. The reproducibility of the scores for some junior radiologists was moderate. This phenomenon may be related to the more stable image reading habits of the senior group during the longer working time frame, which is consistent with the results of Ke et al. (23).

In terms of inter-observer agreement, the present study revealed that the consistency of the T2WI score among senior radiologists was fair and that the consistency among all the readers and the junior group was poor, indicating that the MRI score is related to clinical experience. The consistency of the DWI and PI-RADS categories was poor across all radiologists, with a consistency coefficient of only 0.143 in the senior group, lower than that of T2WI. This phenomenon may be attributed to 2 reasons. On the one hand, the resolution of DWI/ADC images is lower than that of T2WI. DWI is susceptible to interference from magnetic field inhomogeneity and therefore prone to distortion (24). In such cases, accurately defining lesion characteristics, such as signal homogeneity, is difficult. On the other hand, it is also necessary to standardize the postprocessing methods of ADC maps to make the images more homogeneous (25). To solve these problems, DWI image acquisition methods can be developed to improve image quality and enhance lesion identifiability. Furthermore, objective quantitative parameters can be obtained through the development of methods such as artificial intelligence to avoid subjectivity. Other previous studies have reported that the consistency of the PI-RADS category between different radiologists was moderate (23,26). The consistency in this study was lower than that in these studies. The potential reasons for this may be as follows. The radiologists who scored the prostate in their study were all from a single institution, and there was some similarity in image interpretation habits. The present study included readers from institutions at different levels (provincial central hospitals, municipal central hospitals, and county-level hospitals), and the knowledge base of radiologists varied widely. In addition, there was a large variation in the number of years of experience of readers in this study.

Recent artificial intelligence advancements have facilitated large-scale automated analysis of MRI images alongside histopathological findings. Some studies have suggested that the random forest algorithm based on quantitative radiomics features extracted from T2WI, DWI, and DCE images can adjust the PI-RADS score, and the performance of the adjusted PI-RADS was improved (27,28). In another study involving 1,540 MRI scans, the PI-RADS score generated by an artificial intelligence system outperformed more than 70% of general readers in detecting clinically significant PCa (29). Even so, given the potential harm posed by significant prostate lesions, proactive management is essential for individuals with uncertain MRI interpretations. Management strategies extend beyond immediate biopsy, whether random or targeted, to include secondary follow-up imaging for score reassessment and subsequent decision-making (21).

This study has several limitations. First, it only included prostate lesions in the peripheral zone and transition zone but not cases in the central zone or anterior fibrous matrix zone, and thus cannot represent the consistency of all regions. Second, although our readers were from 5 institutions, the MR image data were obtained from 3 machines at a single institution. The differences between the machine and scan parameters may lead to differences in image quality, posing potential challenges to the accuracy and consistency of scoring. Third, to mimic the real clinical scenario, we required the readers to record only the sequences that needed to be scored in the PI-RADS category rather than all the sequences. The number of T2WI-scored cases and DWI scores used for agreement analysis were different. Furthermore, the radiologists were asked to record only the score of the lesion that he/she considered to be the most malignant when multiple lesions in the prostate glands were found. The scoring and categorization of partial patients by different readers may be based on different lesions, which may affect the accuracy of the results. Last but not least, our study included a limited number of 164 cases, and a larger order of magnitude evaluation experiment should be performed to provide more information for improving PI-RADS.

Conclusions

The performance of radiologists with varying experience and working backgrounds to identify prostate lesions via PI-RADS v2.1 was acceptable. However, even among the lesions categorized as 3–4 by the senior radiologist group, the proportion of non-significant lesions was still high, which may lead to many unnecessary punctures. The inter-observer consistency of the PI-RADS sequence and category was poor, and this process was affected by clinical experience. We believe that in subsequent analyses, the resolution of DWI images should be improved, and the consistency of the postprocessing of ADC maps should be specified; meanwhile, the use of methods such as deep learning combined with quantitative quantities may help to avoid subjectivity.

Acknowledgments

We thank Siping Zhu, Huiting Gui, Rui Lu, and Li’e Wu for their contribution to this study by participating as readers.

Footnote

Reporting Checklist: The authors have completed the STARD reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-37/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-37/dss

Funding: This study was funded by Foundation of Jiangxi Province Science and Technology Department (No. 20212ACB206021).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-37/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Second Affiliated Hospital of Nanchang University Medical Research Ethics Committee (No. IIT-0-2024-279) and the requirement for individual consent for this analysis was waived due to the retrospective nature.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, Bray F. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin 2021;71:209-49. [Crossref] [PubMed]
2. Cha EK, Eastham JA. Chemotherapy and novel therapeutics before radical prostatectomy for high-risk clinically localized prostate cancer. Urol Oncol 2015;33:217-25. [Crossref] [PubMed]
3. Kubihal V, Kundra V, Lanka V, Sharma S, Das P, Nayyar R, Das CJ. Prospective evaluation of PI-RADS v2 and quantitative MRI for clinically significant prostate cancer detection in Indian men - East meets West. Arab J Urol 2022;20:126-36. [Crossref] [PubMed]
4. Spratt DE, Zhang J, Santiago-Jiménez M, Dess RT, Davis JW, Den RB, et al. Development and Validation of a Novel Integrated Clinical-Genomic Risk Group Classification for Localized Prostate Cancer. J Clin Oncol 2018;36:581-90. [Crossref] [PubMed]
5. Feng FY, Huang HC, Spratt DE, Zhao SG, Sandler HM, Simko JP, et al. Validation of a 22-Gene Genomic Classifier in Patients With Recurrent Prostate Cancer: An Ancillary Study of the NRG/RTOG 9601 Randomized Clinical Trial. JAMA Oncol 2021;7:544-52. [Crossref] [PubMed]
6. Walker CH, Marchetti KA, Singhal U, Morgan TM. Active surveillance for prostate cancer: selection criteria, guidelines, and outcomes. World J Urol 2022;40:35-42. [Crossref] [PubMed]
7. Chatterjee A, Harmath C, Oto A. New prostate MRI techniques and sequences. Abdom Radiol (NY) 2020;45:4052-62. [Crossref] [PubMed]
8. Turkbey B, Rosenkrantz AB, Haider MA, Padhani AR, Villeirs G, Macura KJ, Tempany CM, Choyke PL, Cornud F, Margolis DJ, Thoeny HC, Verma S, Barentsz J, Weinreb JC. Prostate Imaging Reporting and Data System Version 2.1: 2019 Update of Prostate Imaging Reporting and Data System Version 2. Eur Urol 2019;76:340-51. [Crossref] [PubMed]
9. Weinreb JC, Barentsz JO, Choyke PL, Cornud F, Haider MA, Macura KJ, Margolis D, Schnall MD, Shtern F, Tempany CM, Thoeny HC, Verma S. PI-RADS Prostate Imaging - Reporting and Data System: 2015, Version 2. Eur Urol 2016;69:16-40. [Crossref] [PubMed]
10. Barentsz JO, Richenberg J, Clements R, Choyke P, Verma S, Villeirs G, Rouviere O, Logager V, Fütterer JJEuropean Society of Urogenital Radiology. ESUR prostate MR guidelines 2012. Eur Radiol 2012;22:746-57. [Crossref] [PubMed]
11. Chen F, Cen S, Palmer S. Application of Prostate Imaging Reporting and Data System Version 2 (PI-RADS v2): Interobserver Agreement and Positive Predictive Value for Localization of Intermediate- and High-Grade Prostate Cancers on Multiparametric Magnetic Resonance Imaging. Acad Radiol 2017;24:1101-6. [Crossref] [PubMed]
12. Muller BG, Shih JH, Sankineni S, Marko J, Rais-Bahrami S, George AK, de la Rosette JJ, Merino MJ, Wood BJ, Pinto P, Choyke PL, Turkbey B. Prostate Cancer: Interobserver Agreement and Accuracy with the Revised Prostate Imaging Reporting and Data System at Multiparametric MR Imaging. Radiology 2015;277:741-50. [Crossref] [PubMed]
13. Beetz NL, Haas M, Baur A, Konietschke F, Roy A, Hamm CA, Rudolph MM, Shnayien S, Hamm B, Cash H, Asbach P, Penzkofer T. Inter-Reader Variability Using PI-RADS v2 Versus PI-RADS v2.1: Most New Disagreement Stems from Scores 1 and 2. Rofo 2022;194:852-61. [Crossref] [PubMed]
14. Yan WF, Xu HY, Jiang L, Zhang L, Guo YK, Li Y, Shen LT, Min CY, Yang ZG. Early longitudinal changes in left ventricular function and morphology in diabetic pigs: evaluation by 3.0T magnetic resonance imaging. Cardiovasc Diabetol 2023;22:6. [Crossref] [PubMed]
15. Lee CB, Choi SJ, Ahn JH, Ryu DS, Park MS, Jung SM, Shin DR. Ectopic insertion of the pectoralis minor tendon: inter-reader agreement and findings in the rotator interval on MRI. Korean J Radiol 2014;15:764-70. [Crossref] [PubMed]
16. Van Velthoven MH, Oke J, Kardos A. ChroniSense National Early Warning Score Study: Comparison Study of a Wearable Wrist Device to Measure Vital Signs in Patients Who Are Hospitalized. J Med Internet Res 2023;25:e40226. [Crossref] [PubMed]
17. Thompson JE, Moses D, Shnier R, Brenner P, Delprado W, Ponsky L, Pulbrook M, Böhm M, Haynes AM, Hayen A, Stricker PD. Multiparametric magnetic resonance imaging guided diagnostic biopsy detects significant prostate cancer and could reduce unnecessary biopsies and over detection: a prospective study. J Urol 2014;192:67-74. [Crossref] [PubMed]
18. Grey AD, Chana MS, Popert R, Wolfe K, Liyanage SH, Acher PL. Diagnostic accuracy of magnetic resonance imaging (MRI) prostate imaging reporting and data system (PI-RADS) scoring in a transperineal prostate biopsy setting. BJU Int 2015;115:728-35. [Crossref] [PubMed]
19. Bjurlin MA, Carroll PR, Eggener S, Fulgham PF, Margolis DJ, Pinto PA, Rosenkrantz AB, Rubenstein JN, Rukstalis DB, Taneja SS, Turkbey B. Update of the Standard Operating Procedure on the Use of Multiparametric Magnetic Resonance Imaging for the Diagnosis, Staging and Management of Prostate Cancer. J Urol 2020;203:706-12. [Crossref] [PubMed]
20. Wei X, Xu J, Zhong S, Zou J, Cheng Z, Ding Z, Zhou X. Diagnostic value of combining PI-RADS v2.1 with PSAD in clinically significant prostate cancer. Abdom Radiol (NY) 2022;47:3574-82. [Crossref] [PubMed]
21. Hauth E, Jaeger H, Hohmuth H, Beer M. Follow-up MR imaging of PI-RADS 3 and PI-RADS 4 prostate lesions. Clin Imaging 2017;43:64-8. [Crossref] [PubMed]
22. Chen Y, Lin J, Cao W, Meng T, Ling J, Wen Z, Kong L, Qian L, Guo Y, Zhang W, Wang H. Optimizing prostate biopsy decision-making for patients with Prostate Imaging-Reporting and Data System (PI-RADS) ≥3 lesions: novel magnetic resonance imaging (MRI)-based nomograms. Quant Imaging Med Surg 2024;14:8196-210. [Crossref] [PubMed]
23. Ke Z, Wang L, Min XD, Feng ZY, Kang Z, Zhang PP, Li BS, You HJ, Hou SC. Diagnostic Performance and Interobserver Consistency of the Prostate Imaging Reporting and Data System Version 2: A Study on Six Prostate Radiologists with Different Experiences from Half a Year to 17 Years. Chin Med J (Engl) 2018;131:1666-73. [Crossref] [PubMed]
24. Wang Y, Chen S, Zhang W, Xiao W, Su J, Zhang R, Wei Y, Luo M. Image quality evaluation of diffusion-weighted imaging in bladder cancer: a comparison between integrated slice-specific dynamic shimming and single-shot echo-planar imaging. Quant Imaging Med Surg 2023;13:2526-37. [Crossref] [PubMed]
25. Li J, Xia Y, Dai J, Sun G, Xu M, Lin X, Gu L, Shi J, Liu S, Fan L. Comparison of single-shot, FOCUS single-shot, MUSE, and FOCUS MUSE diffusion weighted imaging for pulmonary lesions: A pilot study. Heliyon 2024;10:e35203. [Crossref] [PubMed]
26. Westphalen AC, McCulloch CE, Anaokar JM, Arora S, Barashi NS, Barentsz JO, et al. Variability of the Positive Predictive Value of PI-RADS for Prostate MRI across 26 Centers: Experience of the Society of Abdominal Radiology Prostate Cancer Disease-focused Panel. Radiology 2020;296:76-84. [Crossref] [PubMed]
27. Bao J, Qiao X, Song Y, Su Y, Ji L, Shen J, Yang G, Shen H, Wang X, Hu C. Prediction of clinically significant prostate cancer using radiomics models in real-world clinical practice: a retrospective multicenter study. Insights Imaging 2024;15:68. [Crossref] [PubMed]
28. Hamm CA, Baumgärtner GL, Biessmann F, Beetz NL, Hartenstein A, Savic LJ, Froböse K, Dräger F, Schallenberg S, Rudolph M, Baur ADJ, Hamm B, Haas M, Hofbauer S, Cash H, Penzkofer T. Interactive Explainable Deep Learning Model Informs Prostate Cancer Diagnosis at MRI. Radiology 2023;307:e222276. [Crossref] [PubMed]
29. Yu R, Jiang KW, Bao J, Hou Y, Yi Y, Wu D, Song Y, Hu CH, Yang G, Zhang YD. PI-RADS(AI): introducing a new human-in-the-loop AI model for prostate cancer diagnosis based on MRI. Br J Cancer 2023;128:1019-29. [Crossref] [PubMed]