A Bimodal Regression-Without-Truth Technique for Quantitative Nuclear Imaging Evaluation

The development of quantitative-imaging (QI)-based biomarkers to guide clinical decision-making, such as identifying patients with vs. without disease, is of strong interest across nuclear medicine applications, motivating the development of numerous QI methods. For clinical translation, there is an important need to assess these methods with patient data. To address the lack of ground truth in clinical settings, no-gold-standard evaluation (NGSE) approaches[1], such as the regression-without-truth (RWT) technique, have been developed. However, existing NGSE approaches primarily assess the accuracy or precision of the quantitative estimates, rather than their ability on the eventual clinical decision-making task. This limitation arises because most NGSE approaches assume that the underlying true quantitative values are drawn from a unimodal distribution, whereas effective biomarkers for binary patient stratification are expected to follow a bimodal distribution. To address this gap, we developed bimodal-RWT (BM-RWT), an NGSE technique for evaluating QI methods based on their ability to stratify patient populations. We validated BM-RWT in the context of evaluating different attenuation compensation (AC) methods[2], including a recently proposed AI-based transmission-less AC method, for dopamine transporter-SPECT (DaT-SPECT) and on the task of distinguishing between patients with normal vs. reduced striatal binding ratio (SBR)[3].