Impact of the size of the normal database on the performance of semi-quantitative analysis in dopamine transporter SPECT

Ziel/Aim This study investigated the impact of the size of the normal database (nDB) on the classification performance of semi-quantitative analysis in dopamine transporter (DAT) SPECT with ¹²³I-FP-CIT in different settings.

Methodik/Methods Subject cohorts comprised 645 subjects from the Parkinson’s Progression Marker Initiative (PPMI), 207 healthy controls (HC) and 438 Parkinson’s disease (PD) patients, and an independent sample of 372 patients from clinical routine patient care, 186 with non-neurodegenerative parkinsonian syndrome (PS) and 186 with neurodegenerative PS. The putaminal specific binding ratio (SBR) was computed using conventional anatomical ROIs predefined in standard space or using hottest voxels (HV) analysis in large predefined ROIs. SBR values were transformed to z-scores using mean and standard deviation of the SBR in the nDB of varying size (n = 5,10,15,…,50). The accuracy for identifying patients with PD or neurodegenerative PS by the cut-off −2.5 on the z-score was used as performance measure. This was repeated for 10,000 randomly selected nDB, separately for each nDB size. Accuracy with the whole set of HC or non-neurodegenerative PS subjects as nDB was used as benchmark.

Ergebnisse/Results Mean loss of accuracy of the putamen SBR z-score was below 1  % when the nDB included at least 15 subjects, independent of the subject sample (PPMI or clinical) and the SBR method (conventional or HV). Variability of the accuracy of the putamen SBR z-score across the 10,000 realizations of the nDB decreased monotonically with increasing nDB size. The loss of accuracy at the 5^th percentile of accuracy estimates was less than 5  % in all settings when the nDB included at least 25 subjects. The benefit of increasing the nDB beyond n = 40 was very small.

Schlussfolgerungen/Conclusions This study suggests that 25 is the minimum size of nDB to reliably achieve good performance of semi-quantitative analysis in DAT SPECT, independent of the setting.