The value of habitat analysis based on F-PSMA-1007 PET/CT images for prostate cancer risk grading.

PURPOSE

To evaluate the performance of habitat analysis by positron emission tomography (PET)/computed tomography (CT) with F-prostate-specific membrane antigen (PSMA)-1007 (F-PSMA-1007 PET/CT) for prediction of risk grading based on the Gleason Score (GS) for primary prostate cancer (PCa).

METHODS

The data of 42 PCa patients who underwent F-PSMA-1007 PET/CT before puncture biopsy or radical prostatectomy were included for analysis. The whole prostate was manually contoured on PET and CT images as the volume of interest (VOI). Using the Otsu algorithm, the VOI was divided into four habitat subregions. Independent risk factors were screened and a combined model was constructed to predict GS grade by univariate logistic regression followed by multivariate logistic regression of habitat (1-4) and clinical factors (SUV, tPSA, fPSA/tPSA, age). Receiver operating characteristic (ROC) curves were drawn and the area under the ROC curve (AUC), sensitivity, and specificity were calculated to evaluate indicator performance. The Kappa consistency test was used to evaluate the agreement between predictive indicators and the model with pathological results. DeLong's test was used to compare the AUC values.

RESULTS

SUV (OR, 1.139; 95% CI, 1.034-1.253; p = 0.008) and the Habitat 2 spatial proportion (OR, 1.166; 95% CI, 1.041-1.307; p = 0.008) were identified by logistic regression analysis as independent risk factors to distinguish the GS grading of PCa, which the Habitat 2 spatial proportion represented the percentage of voxels in the region with PET-high uptake and CT-low density to the VOI. The AUC values of SUV, Habitat 2 spatial proportion, and the combined prediction model were 0.750 (95% CI, 0.597-0.903), 0.716 (95% CI, 0.559-0.873), and 0.823 (95% CI, 0.694-0.951), respectively. The sensitivity of Habitat 2 spatial proportion was 90.91%, which was higher than SUV (72.73%) and the combined model (68.18%). The specificity of the model combining SUV and Habitat 2 spatial proportion for risk classification of PCa was 90.00%, which was higher than either SUV (75.00%) or Habitat 2 spatial proportion (45.00%).

CONCLUSION

The results of this pilot study showed that the combined prediction model, as a non-invasive method, may provide additional value for risk stratification of PCa, offering new perspectives for individualized clinical diagnosis and treatment.

TRIAL REGISTRATION

https://www.chictr.org.cn/ .

TRIAL REGISTRATION

Registration number: ChiCTR2100052238 (retrospectively registered).