Wednesday, December 2 | 1:30 p.m.-2:30 p.m. | W6-SSNMMI06-2 | Room S405
An algorithm that combines PSMA-PET imaging features with routine lab values may help predict which patients with metastatic castration-resistant prostate cancer (mCRPC) will respond to Pluvicto radioligand therapy, according to this presentation.
Presenter Deepak Anand, PhD, of AI developer nference in India, and colleagues developed and validated a multimodal model using pretreatment prostate-specific membrane antigen (PSMA)-PET scans and core laboratory values obtained from the electronic health record (EHR).
The group retrospectively analyzed 302 mCRPC patients who received at least three cycles of Pluvicto and had evaluable pretreatment PSMA-PET scans. Imaging inputs included lesion shape, standardized uptake value (SUV) metrics, and a measure of uptake heterogeneity within tumors. Ten lab parameters, such as PSA and lymphocyte count, were drawn from the EHR.
The combined model distinguished responders from nonresponders with an area under the curve (AUC) of 0.80, outperforming models that used only clinical features (0.73) or only imaging features (0.76). PSA value, lymphocyte count, and SUV mean were among the most predictive features. Responders also had markedly better outcomes, with 30-month survival of about 79% versus 34% for nonresponders.
“These findings support AI-assisted treatment planning for PSMA-targeted radioligand therapy,” the researchers wrote.
Attend this Wednesday afternoon presentation to hear more.
