Voice analysis performed by Machine Learning, a branch of artificial intelligence (A.I.), could be a fundamental tool for the diagnosis and clinical evaluation of Essential Tremor. That is the perspective opened by an Italian study, published in the journal Movement Disorders, resulting from the collaboration between the Department of Human Neurosciences of Sapienza University of Rome, I.R.C.C.S. Neuromed of Pozzilli, Isernia, southern Italy, and the Department of Electronic Engineering of the University of Rome Tor Vergata.

Essential Tremor is the most common movement disorder, affecting 4% of people over the age of 65 and about twenty times more common than Parkinson's disease - with which it is frequently confused. In most cases, it causes an involuntary tremor of the upper limbs, significantly compromising patients' quality of life. About 12% of them also experience a characteristic voice tremor.

Voice is a complex biological phenomenon, requiring the correct activation of an extensive network of neurons in the brain. Accordingly, the voice can provide relevant information on nervous system health. In this context, advanced analysis of voice recordings using up-to-date machine learning algorithms can be considered a new diagnostic frontier.

With its high sensitivity/specificity, A.I. analysis allowed both the automatic recognition of vocal tremor and the monitoring of symptomatic response to specific pharmacological therapies in patients with Essential Tremor. This approach could lead to an innovative and non-invasive tool for the diagnosis of specific neurological diseases.

"The diagnosis of Essential Tremor and the evaluation of the response to drug therapy - says Antonio Suppa of the Department of Human Neurosciences of Sapienza University of Rome and I.R.C.C.S. Neuromed – is currently based on clinical evaluations, which can be affected by the clinician experience. Thanks to our results, we believe it would be possible in the future to develop a new automated and standardized method of voice evaluation."

Still, the outcomes could be even broader, as Professor Suppa explains: "Advanced voice analysis would allow, in the future, to recognize specific neurological disorders and identify the most appropriate therapeutic interventions, in line with the new frontier of precision medicine. Artificial intelligence will also boost the emerging field of telemedicine, opening new perspectives in the remote diagnosis and treatment of neurological disorders."

More in details, the research included voice recording of 58 patients with Essential Tremor and 74 healthy controls. Voice recordings, performed while the participants pronounced a vowel for five seconds, were subsequently examined with machine learning algorithms by Professor Giovanni Costantini and Professor Giovanni Saggio at the laboratories of the Electronic Engineering Department of the University of Rome Tor Vergata.

"The use of artificial intelligence allows us to expand the diagnostic tools currently available in this field, going beyond the current boundaries  - explains Francesco Asci, a neurologist at the Department of Human Neurosciences of Sapienza University of Rome and co-author of the study. In our research, we were able to identify voice alterations even in patients affected by Essential Tremor not manifesting any apparent vocal abnormality on clinical examination. Advanced voice analysis with machine learning algorithms will therefore enable the development of innovative, standardized and high-precision diagnostic procedures."

References:

Voice Analysis with Machine Learning: One Step Closer to an Objective Diagnosis of Essential Tremor - Antonio Suppa, MD, PhD, Francesco Asci, MD, Giovanni Saggio, PhD, Pietro Di Leo, BA, Zakarya Zarezadeh, BA, Gina Ferrazzano, MD, PhD, Giovanni Ruoppolo, MD, Alfredo Berardelli, MD, and Giovanni Costantini, PhD - Movement Disorders https://doi.org/10.1002/mds.28508

Further Information

Antonio Suppa
Department of Human Neurosciences
antonio.suppa@uniroma1.it