Characterization of Effective Half-Life for Instant Single-Time-Point Dosimetry Using Machine Learning

Area of Science:

• Nuclear Medicine
• Medical Imaging
• Radiopharmaceutical Therapy

Background:

• Single-time-point (STP) dosimetry in radiopharmaceutical therapy (RPT) is convenient but limited by strict timing requirements.
• Current STP methods face challenges in routine clinical settings due to late data acquisition needs.

Purpose of the Study:

• To introduce a novel instant STP (iSTP) dosimetry method for RPT.
• To predict organ effective half-life (Teff) using machine learning on pretherapy PET and clinical data.
• To enable flexible and early dosimetry acquisition post-therapy.

Main Methods:

• Developed a machine learning model to predict Teff for kidneys, liver, and spleen using pretherapy PET/CT and clinical data from 22 patients.
• Applied the model to estimate time-integrated activity and absorbed dose for [177Lu]Lu-PSMA I&T RPT.
• Compared iSTP dosimetry results with multiple-time-point and Hänscheid methods using two prediction scenarios.

Main Results:

• The iSTP method allowed early posttherapy time points (2, 20, 43, 69 h) for dosimetry.
• Aggregating 2 and 20 h data in the first scenario yielded mean differences below 27% in time-integrated activity for all organs compared to Hänscheid method.
• No significant differences (P > 0.05) were observed when predicting Teff for subsequent cycles using only initial PET/CT data.

Conclusions:

• Preliminary results demonstrate the feasibility of predicting Teff with pretherapy data for rapid and flexible STP dosimetry post-RPT.
• The proposed iSTP method shows potential to expedite dosimetry application in broader clinical settings, including outpatient RPT.
• This approach could enhance the routine clinical application of personalized dosimetry in RPT.