Correction to: Machine learning classifiers can predict Gleason pattern 4 prostate cancer with greater accuracy than experienced radiologists.

The original version of this article, published on 11 June 2019, unfortunately contained a mistake. The following correction has therefore been made in the original: In section "Multiparametric MRI review," the readers mentioned in the first sentence were partly incorrect.

Machine learning classifiers can predict Gleason pattern 4 prostate cancer with greater accuracy than experienced radiologists.

OBJECTIVE: The purpose of this study was: To test whether machine learning classifiers for transition zone (TZ) and peripheral zone (PZ) can correctly classify prostate tumors into those with/without a Gleason 4 component, and to compare the performance of the best performing classifiers against the opinion of three board-certified radiologists. METHODS: A retrospective analysis of prospectively acquired data was performed at a single center between 2012 and 2015. Inclusion criteria were (i) 3-T mp-MRI compliant with international guidelines, (ii) Likert ≥ 3/5 lesion, (iii) transperineal template ± targeted index lesion biopsy confirming cancer ≥ Gleason 3 + 3. Index lesions from 164 men were analyzed (119 PZ, 45 TZ). Quantitative MRI and clinical features were used and zone-specific machine learning classifiers were constructed. Models were validated using a fivefold cross-validation and a temporally separated patient cohort. Classifier performance was compared against the opinion of three board-certified radiologists. RESULTS: The best PZ classifier trained with prostate-specific antigen density, apparent diffusion coefficient (ADC), and maximum enhancement (ME) on DCE-MRI obtained a ROC area under the curve (AUC) of 0.83 following fivefold cross-validation. Diagnostic sensitivity at 50% threshold of specificity was higher for the best PZ model (0.93) when compared with the mean sensitivity of the three radiologists (0.72). The best TZ model used ADC and ME to obtain an AUC of 0.75 following fivefold cross-validation. This achieved higher diagnostic sensitivity at 50% threshold of specificity (0.88) than the mean sensitivity of the three radiologists (0.82). CONCLUSIONS: Machine learning classifiers predict Gleason pattern 4 in prostate tumors better than radiologists. KEY POINTS: • Predictive models developed from quantitative multiparametric magnetic resonance imaging regarding the characterization of prostate cancer grade should be zone-specific. • Classifiers trained differently for peripheral and transition zone can predict a Gleason 4 component with a higher performance than the subjective opinion of experienced radiologists. • Classifiers would be particularly useful in the context of active surveillance, whereby decisions regarding whether to biopsy are necessitated.

Multi-parametric MRI zone-specific diagnostic model performance compared with experienced radiologists for detection of prostate cancer.

OBJECTIVES: Compare the performance of zone-specific multi-parametric-MRI (mp-MRI) diagnostic models in prostate cancer detection with experienced radiologists. METHODS: A single-centre, IRB approved, prospective STARD compliant 3 T MRI test dataset of 203 patients was generated to test validity and generalisability of previously reported 1.5 T mp-MRI diagnostic models. All patients included within the test dataset underwent 3 T mp-MRI, comprising T2, diffusion-weighted and dynamic contrast-enhanced imaging followed by transperineal template ± targeted index lesion biopsy. Separate diagnostic models (transition zone (TZ) and peripheral zone (PZ)) were applied to respective zones. Sensitivity/specificity and the area under the receiver operating characteristic curve (ROC-AUC) were calculated for the two zone-specific models. Two radiologists (A and B) independently Likert scored test 3 T mp-MRI dataset, allowing ROC analysis for each radiologist for each prostate zone. RESULTS: Diagnostic models applied to the test dataset demonstrated a ROC-AUC = 0.74 (95% CI 0.67-0.81) in the PZ and 0.68 (95% CI 0.61-0.75) in the TZ. Radiologist A/B had a ROC-AUC = 0.78/0.74 in the PZ and 0.69/0.69 in the TZ. Radiologists A and B each scored 51 patients in the PZ and 41 and 45 patients respectively in the TZ as Likert 3. The PZ model demonstrated a ROC-AUC = 0.65/0.67 for the patients Likert scored as indeterminate by radiologist A/B respectively, whereas the TZ model demonstrated a ROC-AUC = 0.74/0.69. CONCLUSION: Zone-specific mp-MRI diagnostic models demonstrate generalisability between 1.5 and 3 T mp-MRI protocols and show similar classification performance to experienced radiologists for prostate cancer detection. Results also indicate the ability of diagnostic models to classify cases with an indeterminate radiologist score. KEY POINTS: • MRI diagnostic models had similar performance to experienced radiologists for classification of prostate cancer. • MRI diagnostic models may help radiologists classify tumour in patients with indeterminate Likert 3 scores.