System based on artificial intelligence can detect Parkinson’s disease

American researchers have developed a system based on artificial intelligence capable of detecting Parkinson’s disease early, pointing out its severity and progression. All this by analyzing the person’s nocturnal breathing.

The relationship between disease and breathing was noted as early as 1817, in the pioneering work of James Parkinson. It was confirmed in later research, but it is still not applied in the evaluation of patients, usually diagnosed on the basis of clinical symptoms related to motor functions, such as tremor and rigidity.

“Parkinson’s disease has motor symptoms, but also non-motor symptoms. Loss of smell and depression, for example, are findings that are often found before the onset of motor symptoms, but as they are very common in the population, it is difficult to attribute them as manifestations. If we start to have other markers, we can add them up to identify the disease early”, explains Carlos Roberto Rieder, president of ABN (Brazilian Academy of Neurology).

For the doctor, the great relevance of the study, led by scientists from MIT (Massachusetts Institute of Technology) and published last week in the scientific journal Nature Medicine, is precisely to assess the ventilatory pattern that may arise before the onset of motor symptoms, which usually occur when there has already been considerable loss of neurons. Early diagnosis is a way to reduce this damage.

“Medical literature has reported several respiratory symptoms of PD [doença de Parkinson]such as respiratory muscle weakness, sleep-disordered breathing, and degeneration in areas of the brain that control breathing, but without our AI-based model, no physician today can detect the disease or assess its severity from breathing,” says the group. responsible for the research.

The scientists created an analysis model using data from 7,671 people, 757 of them with the disease, with a total of more than 120,000 hours of nocturnal respiratory signs. The data were captured in two ways: by a belt placed on the chest or abdomen of the participant that records the breathing throughout the night or by a wireless sensor that, installed in the room, analyzes the radio waves of the environment and extracts the signals. of breath.

In addition to the diagnosis, the biomarker works as an indicator of disease progression, currently performed using a questionnaire called MDS-UPDRS (Movement Disorder Society Unified Parkinson’s Disease Rating Scale). “It is a scale that evaluates various scores given for the presence of tremor and slowness, for example. The fact is that all progression is based on clinical findings”, comments Rieder.

According to the researchers, this clinical assessment is partially subjective, lacks sensitivity to capture small changes in the patient’s condition, and requires frequent visits to health facilities. In addition, clinical trials for approval of new drugs for the disease must last several years, until changes in the MDS-UPDRS can be reported with sufficient statistical confidence.

“In clinical studies, we start using the drug to verify if it alters the evolution of the disease when there is a clinical diagnosis of motor symptoms, but then it ends up being late. So far, there is nothing in this sense, so it is an advance”, says Rieder.

From one of the databases used, the authors of the article compared breathing signals at two visits, six years apart, and found that in 75% of the cases of patients diagnosed at the second meeting, the model was able to predict signs of the disease in the first assessment.

In the research, the scientists verified that the model manages to capture the severity of the disease with statistical significance. It has this capability because it aggregates measurements from several nights in a row, something unfeasible today, since the patient would not be able to go to the clinic several times to repeat the MDS-UPDRS.

“Our approach has the potential to reduce the cost and duration of PD clinical trials and therefore facilitate drug development. The average cost and time to develop PD drugs is approximately $1.3 billion and 13 years, respectively, which limits the interest of many pharmaceutical companies in the search for new therapies”, the authors argue.

They say the system could, in the future, be deployed in the homes of patients with the disease and those at high risk for the disease. Until then, however, the model needs to be tested with all subtypes of the disease and studies with larger samples are needed.