Impact of Clinical History on Maximum PI-RADS Version 2 Score: A Six-Reader 120-Case Sham History Retrospective Evaluation

Intra-practice Urologist-level Variation in Targeted Fusion Biopsy Outcomes

OBJECTIVE

To examine the extent to which the urologist performing biopsy contributes to variation in prostate cancer detection during fusion-guided prostate biopsy.

METHODS

All men in the Michigan Urological Surgery Improvement Collaborative (MUSIC) clinical registry who underwent fusion biopsy at Michigan Medicine from August 2017 to March 2019 were included. The primary outcomes were clinically significant cancer detection rate (defined as Gleason Grade ≥2) in targeted cores and clinically significant cancer detection on targeted cores stratified by PI-RADS score. Bivariate and multivariable logistic regression analyses were performed.

RESULTS

A total of 1133 fusion biopsies performed by 5 providers were included. When adjusting for patient age, PSA, race, family history, prostate volume, clinical stage, and PI-RADS score, there was no significant difference in targeted clinically significant cancer detection rates across providers (range = 38.5%-46.9%, adjusted P-value = .575). Clinically significant cancer detection rates ranged from 11.1% to 16.7% in PI-RADS 3 (unadjusted P = .838), from 24.6% to 43.4% in PI-RADS 4 (adjusted P = .003), and from 69.4% to 78.8% in PI-RADS 5 (adjusted P = .766) lesions.

CONCLUSION

There was a statistically significant difference in clinically significant prostate cancer detection in PI-RADS 4 lesions across providers. These findings suggest that even among experienced providers, variation at the urologist level may contribute to differences in clinically significant cancer detection rates within PI-RADS 4 lesions. However, the relative impact of biopsy technique, radiologist interpretation, and MR acquisition protocol requires further study.

PI-RADSAI: introducing a new human-in-the-loop AI model for prostate cancer diagnosis based on MRI

BACKGROUND

This study aims to develop and validate an artificial intelligence (AI)-aided Prostate Imaging Reporting and Data System (PI-RADSAI) for prostate cancer (PCa) diagnosis based on MRI.

METHODS

The deidentified MRI data of 1540 biopsy-naïve patients were collected from four centres. PI-RADSAI is a two-stage, human-in-the-loop AI capable of emulating the diagnostic acumen of subspecialists for PCa on MRI. The first stage uses a UNet-Seg model to detect and segment biopsy-candidate prostate lesions, whereas the second stage leverages UNet-Seg segmentation is trained specifically with subspecialist' knowledge-guided 3D-Resnet to achieve an automatic AI-aided diagnosis for PCa.

RESULTS

In the independent test set, UNet-Seg identified 87.2% (628/720) of target lesions, with a Dice score of 44.9% (range, 22.8-60.2%) in segmenting lesion contours. In the ablation experiment, the model trained with the data from three centres was superior (kappa coefficient, 0.716 vs. 0.531) to that trained with single-centre data. In the internal and external tests, the triple-centre PI-RADSAI model achieved an overall agreement of 58.4% (188/322) and 60.1% (92/153) with a referential subspecialist in scoring target lesions; when one-point margin of error was permissible, the agreement rose to 91.3% (294/322) and 97.3% (149/153), respectively. In the paired test, PI-RADSAI outperformed 5/11 (45.5%) and matched the performance of 3/11 (27.3%) general radiologists in achieving a clinically significant PCa diagnosis (area under the curve, internal test, 0.801 vs. 0.770, p < 0.01; external test, 0.833 vs. 0.867, p = 0.309).

CONCLUSIONS

Our closed-loop PI-RADSAI outperforms or matches the performance of more than 70% of general readers in the MRI assessment of PCa. This system might provide an alternative to radiologists and offer diagnostic benefits to clinical practice, especially where subspecialist expertise is unavailable.

Performance of Artificial Intelligence-Aided Diagnosis System for Clinically Significant Prostate Cancer with MRI: A Diagnostic Comparison Study

BACKGROUND

The high level of expertise required for accurate interpretation of prostate MRI.

PURPOSE

To develop and test an artificial intelligence (AI) system for diagnosis of clinically significant prostate cancer (CsPC) with MRI.

STUDY TYPE

Retrospective.

SUBJECTS

One thousand two hundred thirty patients from derivation cohort between Jan 2012 and Oct 2019, and 169 patients from a publicly available data (U-Net: 423 for training/validation and 49 for test and TrumpeNet: 820 for training/validation and 579 for test).

FIELD STRENGTH/SEQUENCE

3.0T/scanners, T₂ -weighted imaging (T₂ WI), diffusion-weighted imaging, and apparent diffusion coefficient map.

ASSESSMENT

Close-loop AI system was trained with an Unet for prostate segmentation and a TrumpetNet for CsPC detection. Performance of AI was tested in 410 internal and 169 external sets against 24 radiologists categorizing into junior, general and subspecialist group. Gleason score >6 was identified as CsPC at pathology.

STATISTICAL TESTS

Area under the receiver operating characteristic curve (AUC-ROC); Delong test; Meta-regression I² analysis.

RESULTS

In average, for internal test, AI had lower AUC-ROC than subspecialists (0.85 vs. 0.92, P < 0.05), and was comparable to junior (0.84, P = 0.76) and general group (0.86, P = 0.35). For external test, both AI (0.86) and subspecialist (0.86) had higher AUC than junior (0.80, P < 0.05) and general reader (0.83, P < 0.05). In individual, it revealed moderate diagnostic heterogeneity in 24 readers (Mantel-Haenszel I²  = 56.8%, P < 0.01), and AI outperformed 54.2% (13/24) of readers in summary ROC analysis. In multivariate test, Gleason score, zonal location, PI-RADS score and lesion size significantly impacted the accuracy of AI; while effect of data source, MR device and parameter settings on AI performance is insignificant (P > 0.05).

DATA CONCLUSION

Our AI system can match and to some case exceed clinicians for the diagnosis of CsPC with prostate MRI.

EVIDENCE LEVEL

3 TECHNICAL EFFICACY: Stage 2.

Urinary MyProstateScore (MPS) to Rule out Clinically-Significant Cancer in Men with Equivocal (PI-RADS 3) Multiparametric MRI: Addressing an Unmet Clinical Need

OBJECTIVE

To evaluate the complementary value of urinary MyProstateScore (MPS) testing and multiparametric MRI (mpMRI) and assess outcomes in patients with equivocal mpMRI.

MATERIALS AND METHODS

Included patients underwent mpMRI followed by urine collection and prostate biopsy at the University of Michigan between 2015 -2019. MPS values were calculated from urine specimens using the validated model based on serum PSA, urinary PCA3, and urinary TMPRSS2:ERG. In the PI-RADS 3 population, the discriminative accuracy of PSA, PSAD, and MPS for GG≥2 cancer was quantified by the AUC curve. Decision curve analysis was used to assess net benefit of MPS relative to PSAD.

RESULTS

There were 540 patients that underwent mpMRI and biopsy with MPS available. The prevalence of GG≥2 cancer was 13% for PI-RADS 3, 56% for PI-RADS 4, and 87% for PI-RADS 5. MPS was significantly higher in men with GG≥2 cancer [median 44.9, IQR (29.4 -57.5)] than those with negative or GG1 biopsy [median 29.2, IQR (14.8 -44.2); P <.001] in the overall population and when stratified by PI-RADS score. In the PI-RADS 3 population (n = 121), the AUC for predicting GG≥2 cancer was 0.55 for PSA, 0.62 for PSAD, and 0.73 for MPS. MPS provided the highest net clinical benefit across all pertinent threshold probabilities.

CONCLUSION

In patients that underwent mpMRI and biopsy, MPS was significantly associated with GG≥2 cancer across all PI-RADS scores. In the PI-RADS 3 population, MPS significantly outperformed PSAD in ruling out GG≥2 cancer. These findings suggest a complementary role of MPS testing in patients that have undergone mpMRI.

Metastatic Versus Osteoporotic Vertebral Fractures on MRI: A Blinded, Multicenter, and Multispecialty Observer Agreement Evaluation

BACKGROUND

MRI is assumed to be valid for distinguishing metastatic vertebral fractures (MVFs) from osteoporotic vertebral fractures (OVFs). This study assessed (1) concordance between the image-based diagnosis of MVF versus OVF and the reference (biopsy or follow-up of >6 months), (2) interobserver and intraobserver agreement on key imaging findings and the diagnosis of MVF versus OVF, and (3) whether disclosing a patient's history of cancer leads to variations in diagnosis, concordance, or agreement.

PATIENTS AND METHODS

This retrospective cohort study included clinical data and imaging from 203 patients with confirmed MVF or OVF provided to 25 clinicians (neurosurgeons, radiologists, orthopedic surgeons, and radiation oncologists). From January 2018 through October 2018, the clinicians interpreted images in conditions as close as possible to routine practice. Each specialist assessed data twice, with a minimum 6-week interval, blinded to assessments made by other clinicians and to their own previous assessments. The kappa statistic was used to assess interobserver and intraobserver agreement on key imaging findings, diagnosis (MVF vs OVF), and concordance with the reference. Subgroup analyses were based on clinicians' specialty, years of experience, and complexity of the hospital where they worked.

RESULTS

For diagnosis of MVF versus OVF, interobserver agreement was fair, whereas intraobserver agreement was substantial. Only the latter improved to almost perfect when a patient's history of cancer was disclosed. Interobserver agreement for key imaging findings was fair or moderate, whereas intraobserver agreement on key imaging findings was moderate or substantial. Concordance between the diagnosis of MVF versus OVF and the reference was moderate. Results were similar regardless of clinicians' specialty, experience, and hospital category.

CONCLUSIONS

When MRI is used to distinguish MVF versus OVF, interobserver agreement and concordance with the reference were moderate. These results cast doubt on the reliability of basing such a diagnosis on MRI in routine practice.

18F-Choline PET/mpMRI for Detection of Clinically Significant Prostate Cancer: Part 1. Improved Risk Stratification for MRI-Guided Transrectal Prostate Biopsies

A prospective single-arm clinical trial was conducted to determine whether ¹⁸F-choline PET/mpMRI can improve the specificity of multiparametric MRI (mpMRI) of the prostate for Gleason ≥ 3+4 prostate cancer. Methods: Before targeted and systematic prostate biopsy, mpMRI and ¹⁸F-choline PET/CT were performed on 56 evaluable subjects with 90 Likert score 3-5 mpMRI target lesions, using a ¹⁸F-choline target-to-background ratio of greater than 1.58 to indicate a positive ¹⁸F-choline result. Prostate biopsies were performed after registration of real-time transrectal ultrasound with T2-weighted MRI. A mixed-effects logistic regression was applied to measure the performance of mpMRI (based on prospective Likert and retrospective Prostate Imaging Reporting and Data System, version 2 [PI-RADS], scores) compared with ¹⁸F-choline PET/mpMRI to detect Gleason ≥ 3+4 cancer. Results: The per-lesion accuracy of systematic plus targeted biopsy for mpMRI alone was 67.8% (area under receiver-operating-characteristic curve [AUC], 0.73) for Likert 4-5 and 70.0% (AUC, 0.76) for PI-RADS 3-5. Several PET/MRI models incorporating ¹⁸F-choline with mpMRI data were investigated. The most promising model selected all high-risk disease on mpMRI (Likert 5 or PI-RADS 5) plus low- and intermediate-risk disease (Likert 4 or PI-RADS 3-4), with an elevated ¹⁸F-choline target-to-background ratio greater than 1.58 as positive for significant cancer. Using this approach, the accuracy on a per-lesion basis significantly improved to 88.9% for Likert (AUC, 0.90; P < 0.001) and 91.1% for PI-RADS (AUC, 0.92; P < 0.001). On a per-patient basis, the accuracy improved to 92.9% for Likert (AUC, 0.93; P < 0.001) and to 91.1% for PI-RADS (AUC, 0.91; P = 0.009). Conclusion: ¹⁸F-choline PET/mpMRI improved the identification of Gleason ≥ 3+4 prostate cancer compared with mpMRI, with the principal effect being improved risk stratification of intermediate-risk mpMRI lesions.

Radiographic stool quantification: an equivalence study of 484 symptomatic and asymptomatic subjects

PURPOSE

To determine if symptomatic patients referred for radiographic stool quantification have equivalent stool burden to asymptomatic patients.

METHOD

This was an IRB-approved HIPAA-compliant retrospective equivalence cohort study. An a priori equivalence power calculation was performed. Consecutive abdominal radiographs performed in adult outpatients with bloating, constipation, diarrhea, or abdominal pain to assess "fecal loading" [n = 242 (fecal cohort)] were compared to those performed in asymptomatic adult outpatients to assess "renal stones" [n = 242 (renal cohort)]. Radiographs were randomized and reviewed by two blinded independent abdominal radiologists. Exclusion criteria, designed to avoid unblinding, included urinary tract calculi ≥ 0.5 cm, multiple urinary tract calculi, and ureteral stent(s). Readers scored all radiographs (n = 484) for stool burden using validated Leech criteria [scale: 0 (none) to 15 (extreme diffuse)]. Mean Leech scores and 95% confidence intervals were calculated. Multivariable generalized linear modeling was performed to adjust for baseline medication use, age, and gender. The adjusted parameter estimate was used to test for equivalence in the mean difference between cohorts using Schuirmann's method of two one-sided t-tests. Inter-reader agreement was assessed with intraclass correlation coefficients.

RESULTS

Overall mean Leech scores for fecal [6.9 (95% CI 6.7, 7.2)] and renal [7.3 (95% CI 7.1, 7.5)] cohorts were equivalent within a margin of 0.75 (adjusted mean difference: - 0.4 [90% CI - 0.7, - 0.04]; p value = 0.02). Inter-reader agreement was good [ICC: 0.62 (95% CI 0.56, 0.68)].

CONCLUSION

Radiographic stool quantification produces equivalent results in symptomatic and asymptomatic adults and is of uncertain value.