Artificial-intelligence-based MRI brain volumetry in patients with essential tremor and tremor-dominant Parkinson's disease

Abstract

Essential tremor and Parkinson's disease patients may present with various tremor types. Overlapping tremor features can be challenging to diagnosis and misdiagnosis is common. Although underlying neurodegenerative mechanisms are suggested, neuroimaging studies arrived at controversial results and often the different tremor types were not considered. We investigated whether different tremor types displayed distinct structural brain features. Structural MRI of 61 patients with essential tremor and 29 with tremor-dominant Parkinson's disease was analysed using a fully automated artificial-intelligence-based brain volumetry to compare volumes of several cortical and subcortical regions. Furthermore, essential tremor subgroups with and without rest tremor or more pronounced postural and kinetic tremor were investigated. Deviations from an internal reference collective of age- and sex-adjusted healthy controls and volumetric differences between groups were examined; regression analysis was used to determine the contribution of disease-related factors on volumetric measurements. Compared with healthy controls, essential tremor and tremor-dominant Parkinson's disease patients displayed deviations in the occipital lobes, hippocampus, putamen, pallidum and mesencephalon while essential tremor patients exhibited decreased volumes within the nucleus caudatus and thalamus. Analysis of covariance revealed similar volumetric patterns in both diseases. Essential tremor patients without rest tremor showed a significant atrophy within the thalamus compared to tremor-dominant Parkinson's disease and atrophy of the mesencephalon and putamen were found in both subgroups compared to essential tremor with rest tremor. Disease-related factors contribute to volumes of occipital lobes in both diseases and to volumes of temporal lobes in essential tremor and the putamen in Parkinson's disease. Fully automated artificial-intelligence-based volumetry provides a fast and rater-independent method to investigate brain volumes in different neurological disorders and allows comparisons with an internal reference collective. Our results indicate that essential tremor and tremor-dominant Parkinson's disease share structural changes, indicative of neurodegenerative mechanisms, particularly of the basal-ganglia-thalamocortical circuitry. A discriminating, possibly disease-specific involvement of the thalamus was found in essential tremor patients without rest tremor and the mesencephalon and putamen in tremor-dominant Parkinson's disease and essential tremor without rest tremor.

Keywords: Parkinson’s disease; artificial intelligence; brain volumetry; tremor.

Graphical abstract

Figure 1

Percentage proportion of subjects with volume deviations (%). Incidence of deviations [two or four standard deviations] of the single brain volumes compared with a norm collective. Deviations of supratentorial volumes are provided separately for each side only (the volume contralateral to the clinically more affected side has been indicated with *). Deviations of cortical and subcortical volumes correspond to atrophy, deviations of ventricular volumes to enlargement. Significant deviations were found within the nucleus caudatus contralateral to the clinically more affected side and the thalamus contralateral to the clinically less affected side.

Figure 2

Scatter plots showing volumes correlating with disease-related factors (disease duration or tremor severity score CRST) while controlled for age and sex using hierarchical multiple linear regression. In ET, significant correlations (⁑) were found for the TBV (ΔF = 5.25, ΔR² = 0.09) with disease duration (P = 0.01) and CRST scores (P = 0.04). Furthermore, significant effects were evident for the GM (ΔF = 3.6, ΔR² = 0.08) with CRST score (P = 0.01). All volumes have been normalized to the ICV.

Figure 3

Scatter plots showing volumes correlating with disease-related factors (disease duration or tremor severity score CRST) while controlled for age and sex using hierarchical multiple linear regression. In ET, significant correlations (⁑) were found for the lateral ventricles (ΔF = 5.0, ΔR² = 0.09) with disease duration (P = 0.02) and CRST scores (P = 0.047). Furthermore, significant effects were evident for the temporal lobe (ΔF = 4.37, ΔR² = 0.13) with CRST score (P = 0.005) and the occipital lobe with disease duration (ΔF = 4.22, ΔR² = 0.13, P = 0.006). In td Parkinson’s disease, disease duration contributed to occipital lobe volume (ΔF = 4.03, ΔR² = 0.24, P = 0.01) and the CRST to the putaminal volume (ΔF = 6.84, ΔR² = 0.37, P = 0.005). The volume contralateral to the clinically more affected side has been indicated with *. All volumes have been normalized to the ICV.