Multi-institutional Clinical Tool for Predicting High-risk Lesions on 3Tesla Multiparametric Prostate Magnetic Resonance Imaging

The Low Rate of Adherence to Checklist for Artificial Intelligence in Medical Imaging Criteria Among Published Prostate MRI Artificial Intelligence Algorithms

OBJECTIVE

To determine the rigor, generalizability, and reproducibility of published classification and detection artificial intelligence (AI) models for prostate cancer (PCa) on MRI using the Checklist for Artificial Intelligence in Medical Imaging (CLAIM) guidelines, a 42-item checklist that is considered a measure of best practice for presenting and reviewing medical imaging AI research.

MATERIALS AND METHODS

This review searched English literature for studies proposing PCa AI detection and classification models on MRI. Each study was evaluated with the CLAIM checklist. The additional outcomes for which data were sought included measures of AI model performance (eg, area under the curve [AUC], sensitivity, specificity, free-response operating characteristic curves), training and validation and testing group sample size, AI approach, detection versus classification AI, public data set utilization, MRI sequences used, and definition of gold standard for ground truth. The percentage of CLAIM checklist fulfillment was used to stratify studies into quartiles. Wilcoxon's rank-sum test was used for pair-wise comparisons.

RESULTS

In all, 75 studies were identified, and 53 studies qualified for analysis. The original CLAIM items that most studies did not fulfill includes item 12 (77% no): de-identification methods; item 13 (68% no): handling missing data; item 15 (47% no): rationale for choosing ground truth reference standard; item 18 (55% no): measurements of inter- and intrareader variability; item 31 (60% no): inclusion of validated interpretability maps; item 37 (92% no): inclusion of failure analysis to elucidate AI model weaknesses. An AUC score versus percentage CLAIM fulfillment quartile revealed a significant difference of the mean AUC scores between quartile 1 versus quartile 2 (0.78 versus 0.86, P = .034) and quartile 1 versus quartile 4 (0.78 versus 0.89, P = .003) scores. Based on additional information and outcome metrics gathered in this study, additional measures of best practice are defined. These new items include disclosure of public dataset usage, ground truth definition in comparison to other referenced works in the defined task, and sample size power calculation.

CONCLUSION

A large proportion of AI studies do not fulfill key items in CLAIM guidelines within their methods and results sections. The percentage of CLAIM checklist fulfillment is weakly associated with improved AI model performance. Additions or supplementations to CLAIM are recommended to improve publishing standards and aid reviewers in determining study rigor.

Integration of magnetic resonance imaging into prostate cancer nomograms

The decision whether to undergo prostate biopsy must be carefully weighed. Nomograms have widely been utilized as risk calculators to improve the identification of prostate cancer by weighing several clinical factors. The recent inclusion of multiparametric magnetic resonance imaging (mpMRI) findings into nomograms has drastically improved their nomogram's accuracy at identifying clinically significant prostate cancer. Several novel nomograms have incorporated mpMRI to aid in the decision-making process in proceeding with a prostate biopsy in patients who are biopsy-naïve, have a prior negative biopsy, or are on active surveillance. Furthermore, novel nomograms have incorporated mpMRI to aid in treatment planning of definitive therapy. This literature review highlights how the inclusion of mpMRI into prostate cancer nomograms has improved upon their performance, potentially reduce unnecessary procedures, and enhance the individual risk assessment by improving confidence in clinical decision-making by both patients and their care providers.

Multiparametric MRI Versus SelectMDx Accuracy in the Diagnosis of Clinically Significant PCa in Men Enrolled in Active Surveillance

BACKGROUND/AIM

To evaluate the diagnostic accuracy of the urinary SelectMDx test in the diagnosis of clinically significant prostate cancer (csPCa) in men enrolled in an active surveillance (AS) protocol.

PATIENTS AND METHODS

From July 2015 to July 2018, 125 men with very low-risk PCa were enrolled in the AS protocol; all patients underwent confirmatory transperineal saturation biopsy (SPBx). In the presence of PI-RADS score ≥3, a targeted MRI/TRUS fusion-guided biopsy was added to SPBx. Post-digital rectal examination urine was collected in 45/125 (36%) patients before SPBx; the genetic urine analysis was performed using a biomarker-based risk score model, the SelectMDx, that measured mRNA levels of distal-less homeobox 1 (DLX1) and homeobox C6 (HOXC6).

RESULTS

A total of 9/45 (20%) patients were reclassified as csPCa (7 cases=Grade Group 2; 2 cases=Grade Group 3); sensitivity, specificity, positive predictive value, negative predictive value and diagnostic accuracy of mpMRI vs. SelectMDx in the diagnosis of csPCa were equal to 66.6 vs. 55.6%, 87.7 vs. 65.8%, 54.5 vs. 27.8%, 92.3 vs. 87%, 84.9 vs. 70.3%, respectively.

CONCLUSION

SPBx combined with MRI/TRUS fusion biopsy significantly outperformed the diagnostic accuracy of SelectMDx (70.3%) in the diagnosis of csPCa in men enrolled in AS.

Cribriform pattern and perineural invasion on MR/US fusion biopsy predict failure of selection criteria for prostatic hemigland ablation

OBJECTIVES

To assess clinicopathologic factors on MR/US fusion biopsy that might predict failure of theoretical selection criteria for prostatic hemigland ablation (HA).

SUBJECTS AND METHODS

A retrospectively maintained single institution multiparametric MRI database (n = 1667) was queried to identify 355 patients who underwent MR/US fusion biopsy, including both targeted biopsy and concurrent systematic biopsy from December 1, 2014 to June 1, 2018. Clinical, pathological, and imaging variables were assessed on fusion biopsy (Table 1) to determine who met theoretical selection criteria for HA, defined as unilateral intermediate-risk prostate cancer per NCCN criteria (Grade Group [GG] 2 or 3 with prostate-specific antigen <20) and no evidence of extraprostatic extension (EPE) on multiparametric MRI. Predictors of selection criteria failure were then assessed in patients who also underwent radical prostatectomy (RP). Failure of the theoretical HA selection criteria was defined as presence of GG ≧ 2 on the contralateral (untreated) side, or the presence of high-risk disease (any GG ≧ 4 or EPE) in the RP specimen.

RESULTS

Of the 355 patients who underwent fusion biopsy, 84 patients met the theoretical selection criteria for HA. Of those patients eligible, 54 underwent RP, 37 (68.5%) of which represented unsuccessful HA selection criteria. Patients no longer met HA selection criteria on the basis of upgrading alone in 6/54 (11.1%), EPE alone in 9/54 (16.7%), bilateral GG 2 or 3 in 16/54 (29.6%) or combined EPE and bilateral GG 2 or 3 in 6/54 (11.1%) cases. In the HA selection failures due to upgrading, three also had EPE, one of whom also had missed contralateral GG ≧ 2 disease. The only factor independently associated with HA failure was any presence of cribriform pattern (HR 7.01, P = 0.021). Perineural invasion on systematic biopsyalso appeared to improve the performance of our multivariable model (HR 5.33, P = 0.052), though it was not statistically significant when using a cutoff of <0.05. Accuracy for predicting successful HA was 0.32 and improved to 0.74 if PNI or cribriform were excluded and 0.84 if both were excluded.

CONCLUSIONS

In a retrospective analysis of RP patients who underwent preoperative MRI/US fusion biopsy, current selection criteria for prostatic HA based on NCCN intermediate-risk stratification failed to accurately identify appropriate candidates in 68.5% of patients. Cribriform pattern and PNI detected on biopsy reduced the failure of hemigland selection criteria to 43%. These criteria should be routinely reported on biopsy pathology and taken into consideration when selecting patients for HA in prospective clinical trials.