Transcranial Magnetic Stimulation Accurately Differentiates Dementias
Researchers developed and trained a random forest classifier that resulted in high precision and accuracy as a screening tool for those with mild cognitive impairment.
Alberto Benussi, MD
A recent study evaluated a random forest (RF) classifier on transcranial magnetic stimulation (TMS) measures in patients with mild cognitive impairment (MCI) and found TMS to be a useful and accurate screening tool for MCI.
Researchers employed a series of 3 binary classifiers which all resulted in the prediction model exhibiting high classification accuracy (range, 0.72–0.86), high precision (range, 0.72–0.90), high recall (range, 0.75–0.98) and high F1-scores (range, 0.78–0.92) in differentiation of neurodegenerative disorders. Classification indices showed higher accuracy (range, 0.83–0.93), precision (range, 0.87–0.89), recall (range, 0.83–1.00), and F1 scores (range, 0.85–0.94) when using a new classifier trained and validated on the cohort of patients with MCI.
First author Alberto Benussi, MD, neurologist, Department of Clinical and Experimental Sciences, University of Brescia, Italy, and colleagues wrote that “our group has recently developed an index using [TMS] measures of intracortical circuit excitability. TMS allows to non-invasively assess neurotransmitters imbalance and it has been demonstrated helpful to differentiate AD, [frontotemporal dementia] and [dementia with Lewy bodies] with high accuracy.”
Benussi and colleagues analyzed data from 153 participants, 64 with MCI-Alzheimer Disease (MCI-AD), 28 with MCI-frontotemporal dementia (MCI-FTD), 14 with MCI-Lewy body dementia (MCI-DLB), and 47 healthy controls. TMS measures assessed included short latency afferent inhibition (SAI), short interval intracortical inhibition and facilitation (SICI-ICF), and long interval intracortical inhibition (LICI).
READ MORE: P-Tau181 Associated With Longitudinal Neurodegeneration in Alzheimer Disease
It has been previously shown that AD and DLB are characterized by a deficit in SAI while FTD and DLB show a prominent change in SICI-IF.
At least 1 amyloid marker (PET amyloid or CSF Aβ1-42) was seen in 61 out of 106 (57.7%) patients with MCI and at least 1 marker of neuronal injury (CSF Tau or 18F-FDG-PET or SPECT-DaTSCAN) was seen in 87 (82.1%) of patients with MCI. Out of 28 patients with MCI-FTD, 8 (28.6%) had an inherited monogenic disorder (4 GRN, 3 C9orf72, and 1 MAPT mutation). An at least 18-month follow-up was performed in 43 (40.6%) patients with MCI, with follow-up performed in an average of 23.1 months (standard deviation [SD], 12.7).
Analysis of covariance (ANCOVA) revealed significant interaction for SICI-ICF [F(12.5, 392.3), 14.0; P <.001; partial η2, 0.31; ε, 0.70] and SAI [F(7.1, 220.0), 2.9; P =.002; partial η2, 0.09; ε = 0.79]. No significant interaction was observed between inter-stimulus interval (ISI) and diagnosis [F(4.6, 85.4), 0.9; P =.494; partial η2, 0.05; ε = 0.76] and a simple main effect for ISI [F(1.5, 85.4), 4.0; P =.031; partial η2, 0.07; ε, 0.76]. SICI-ICF results were significantly impaired in both MCI-FTD and MCI-DLB, while SAI was significantly impaired in MCI-AD and MCI-DLB.
No significant associations were observed when evaluating MCI-AD between CSF measures and TMS measures (all P >.05).
“The addition of TMS measures to the routine diagnostic assessment could allow for an earlier diagnosis, when combined with clinical and conventional methods of diagnosis. These findings support for the use of TMS intracortical excitability measures to be translated from the experimental setting to the clinical practice, even in the prodromal phases of neurodegenerative dementias,” Benussi and colleagues concluded.
