Utilizing Deep Learning to Automate TTV Generation on 177Lu-PSMA SPECT/CT

The utilization of total tumor volume (TTV) and whole-body mean standardized uptake value (SUV_mean) on ¹⁷⁷Lu-PSMA single-photon emission computed tomography/computed tomography (SPECT/CT; Lu-SPECT) can predict outcomes in patients with metastatic castration-resistant prostate cancer (mCRPC) who are receiving ¹⁷⁷Lu-PSMA therapy. As TTV is a lengthy process, researchers have examined the efficacy of utilizing an algorithm developed by deep learning to automatically generate TTV on Lu-SPECT.

A total of 95 ¹⁷⁷Lu-PSMA therapy SPECT/CT scans from 36 patients with mCRPC were gathered from the LuPIN trial, which analyzed the safety and efficacy of ¹⁷⁷Lu-PSMA 617 and idronoxil (NOX66) in patients with end-stage mCRPC. Each patient was administered 1 to 6 cycles of therapy, with each cycle occurring every 6 weeks. The algorithm segmented all volumes of interest, with a 3-SUV threshold. The proposed method utilized CT-based, deep learning-generated normal organs for automatic removal of physiologic uptake on Lu-SPECT.

The average TTV for the proposed method was 927 mL compared with an average Gleason score (GS) TTV of 847 mL, which was not deemed significantly different (P=.054). The average absolute difference in TTV between the proposed method and GS was 110 mL.

The average SUV_mean of the proposed method was 8.60, and the average SUV_mean of GS was 8.96. This difference was deemed significant (P=.03). The average relative error in SUV_mean was 4.7%, and the median relative error was 1.4%.

The proposed method of algorithm-based TTV produced comparable TTV and whole-body SUV_mean levels. A small but statistically significant difference in SUV_mean was noted between the proposed method and GS, but the clinical significance of this difference is unclear.

A tool using the proposed method may prove to be beneficial by increasing efficacy and reproducibility when generating TTV and whole-body SUV_mean for the assessment of ¹⁷⁷Lu-PSMA therapy. Further research is needed to assess the actual amount of time saved when utilizing automatic physiological uptake removal.