Multivariable Prediction Model for Prostate Cancer Combining PET/CT and MRI

1305 consecutive patients with suspected prostate cancer underwent both mpMRI and 18F-PSMA PET/CT with biopsy at Peking University First Hospital. The dataset was used to develop and validate a multivariable model for predicting clinically significant prostate cancer. The final model, authored by Chaojian Yu and last updated in May 2026, retained three predictors: PRIMARY score, PI-RADS score, and PSAD.

Use Cases

• Training a risk stratification model based on combined imaging parameters PRIMARY score, PI-RADS score, and PSAD.
• Validating the performance of a diagnostic prediction model using temporal validation cohorts.
• Comparing the clinical utility of a multivariable model against individual imaging parameters like mpMRI or PSMA PET/CT.
• Optimizing biopsy decision protocols using a model with reported sensitivity of 96.0% at a specific cutoff.

Strengths

• Model was developed and validated on a cohort of 1305 patients.
• The final model achieved an AUC of 0.916 in training and 0.837 in temporal validation.
• Model performance metrics are detailed, including sensitivity (96.0%) and specificity (76.6%) at defined cutoffs.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data reflects a single-institution study from Peking University First Hospital, which may limit generalizability.

Provenance

Source

Peking University First Hospital

Collection Method

Retrospective study of consecutive patients who underwent mpMRI, 18F-PSMA PET/CT, and biopsy.

Freshness

Last updated 2026-05-18 05:41:27; freshness should be verified.

Geography

China (inferred from hospital location)