A Machine Learning–Based Approach to Discrimination of Tauopathies Using [¹⁸F]PM‐PBB3 PET Images

Abstract

A BS TRACT: Background: We recently developed a positron emission tomography (PET) probe, [ 18 F]PM-PBB3, to detect tau lesions in diverse tauopathies, including mixed three-repeat and four-repeat (3R + 4R) tau fibrils in Alzheimer's disease (AD) and 4R tau aggregates in progressive supranuclear palsy (PSP). For wider availability of this technology for clinical settings, biasfree quantitative evaluation of tau images without a priori disease information is needed. Objective: We aimed to establish tau PET pathology indices to characterize PSP and AD using a machine learning approach and test their validity and tracer capabilities. Methods: Data were obtained from 50 healthy control subjects, 46 patients with PSP Richardson syndrome, and 37 patients on the AD continuum. Tau PET data from 114 regions of interest were subjected to Elastic Net cross-validation linear classification analysis with a oneversus-the-rest multiclass strategy to obtain a linear function that discriminates diseases by maximizing the area under the receiver operating characteristic curve.We defined PSP-and AD-tau scores for each participant as values of the functions optimized for differentiating PSP (4R) and AD (3R + 4R), respectively, from others. Results: The discriminatory ability of PSP-and AD-tau scores assessed as the area under the receiver operating characteristic curve was 0.98 and 1.00, respectively. PSPtau scores correlated with the PSP rating scale in patients with PSP, and AD-tau scores correlated with Mini-Mental State Examination scores in healthy control-AD continuum patients. The globus pallidus and amygdala were highlighted as regions with high weight coefficients for determining PSP-and AD-tau scores, respectively. Conclusions: These findings highlight our technology's unbiased capability to identify topologies of 3R + 4R versus 4R tau deposits.

Radiation dosimetry and pharmacokinetics of the tau PET tracer florzolotau (18F) in healthy Japanese subjects

“…Florzolotau is a new chemical entity discovered at Institutes for Quantum Life and Medical Science at National Institutes for Quantum Science and Technology, Chiba, Japan. Results from nonclinical and clinical studies so far have shown that florzolotau binds to tau lesions with high affinity and less non-specific off-target binding including MAO-B and MAO-A, and shown the probe captures not only Alzheimer's disease but also non-Alzheimer-type tau pathologies with a dynamic range sufficient for differentiation [13,14], and these findings have been confirmed by other investigators [15]. The binding sites of florzolotau in the β-helix of paired helical and straight filaments of tau from Alzheimer's disease have been reported [16].…”

Development and Performance Assessment of a Novel Plasma p-Tau181 Assay Reflecting Tau Tangle Pathology in Alzheimer’s Disease

“…To validate our hypothesis, we collected plasma samples from subjects who underwent PET with both 11 C-Pittsburgh Compound B ( 11 C-PiB) and 18 F-florzorotau (aka PM-PBB3/APN-1607) for the examination of Aβ and tau depositions in the brain, respectively. 34,35 We demonstrated a strong correlation of between plasma levels of p-tau181 fragments measured using our mid-region-directed assay (hereinafter called mid-p-tau181) and tau PET tracer retention but no linear correlation between plasma N-p-tau181 levels and tau PET data, suggesting the clinical usefulness of the novel assay as a surrogate biomarker for PET-visible tau pathologies.…”

Clinical and neuroimaging features of the progressive supranuclear palsy- corticobasal degeneration continuum

“…Visual assessment of PI-2620 images in one study provided better differentiation of PSP-RS from healthy controls compared with visual reads of MRI, and an index considering quantitative values from both PI-2620 (globus pallidus uptake) and MRI (Midbrain-to-pons ratio) performed better than the midbrain-to-pons ratio alone in differentiating PSP and controls (AUROC = 0.98 versus 0.93), particularly in clinically less affected patients [20 ▪ ]. A machine learning approach was also utilized to demonstrate that PM-PBB3 could differentiate PSP-RS from Alzheimer's dementia with high sensitivity (87.6%) and specificity (98.6%), with the globus pallidus and medial temporal lobe the most important regions for differentiation [21 ▪ ]. PM-PBB3 can also differentiate PSP-RS from PD and MSA [18].…”