Classification Accuracy of Transcranial Magnetic Stimulation for the Diagnosis of Neurodegenerative Dementias

• Published in: Annals of neurology Vol. 87; no. 3; pp. 394 - 404
• Main Authors: Benussi, Alberto, Grassi, Mario, Palluzzi, Fernando, Koch, Giacomo, Di Lazzaro, Vincenzo, Nardone, Raffaele, Cantoni, Valentina, Dell'Era, Valentina, Premi, Enrico, Martorana, Alessandro, Lorenzo, Francesco, Bonnì, Sonia, Ranieri, Federico, Capone, Fioravante, Musumeci, Gabriella, Cotelli, Maria Sofia, Padovani, Alessandro, Borroni, Barbara
• Format: Journal Article
• Language: English
• Published: Hoboken, USA John Wiley & Sons, Inc 01.03.2020; Wiley Subscription Services, Inc
• Subjects: Accuracy; Alzheimer's disease; Classification; Classifiers; Dementia; Dementia disorders; Diagnosis; Diagnostic systems; Differential diagnosis; Disorders; Frontotemporal dementia; Lewy bodies; Magnetic fields; Model accuracy; Neurodegenerative diseases; Prediction models; Recall; Transcranial magnetic stimulation
• DOI: 10.1002/ana.25677
• ISSN: 0364-5134

Objective Transcranial magnetic stimulation (TMS) has been suggested as a reliable, noninvasive, and inexpensive tool for the diagnosis of neurodegenerative dementias. In this study, we assessed the classification performance of TMS parameters in the differential diagnosis of common neurodegenerative disorders, including Alzheimer disease (AD), dementia with Lewy bodies (DLB), and frontotemporal dementia (FTD). Methods We performed a multicenter study enrolling patients referred to 4 dementia centers in Italy, in accordance with the Standards for Reporting of Diagnostic Accuracy. All patients underwent TMS assessment at recruitment (index test), with application of reference clinical criteria, to predict different neurodegenerative disorders. The investigators who performed the index test were masked to the results of the reference test and all other investigations. We trained and tested a random forest classifier using 5‐fold cross‐validation. The primary outcome measures were the classification accuracy, precision, recall, and F1 score of TMS in differentiating each neurodegenerative disorder. Results A total of 694 participants were included, namely 273 patients diagnosed as AD, 67 as DLB, and 207 as FTD, and 147 healthy controls (HC). A series of 3 binary classifiers was employed, and the prediction model exhibited high classification accuracy (ranging from 0.89 to 0.92), high precision (0.86–0.92), high recall (0.93–0.98), and high F1 scores (0.89–0.95) in differentiating each neurodegenerative disorder. Interpretation TMS is a noninvasive procedure that reliably and selectively distinguishes AD, DLB, FTD, and HC, representing a useful additional screening tool to be used in clinical practice. Ann Neurol 2020;87:394–404