How speech patterns and tone can help diagnose neurological diseases

Clues in how you speak can indicate health problems long before a patient is officially diagnosed. These clues include the words you choose, how fast you talk, how long you pause, and even tiny derivations and wobbles from regular speech that are too subtle for the human ear to note.

One significant illness with symptoms of speech impairment is dementia. Dementia is a neurological condition that interferes with one’s cognitive functioning, impacting your thinking, remembering, and reasoning to the extent that it interferes with your daily life. When it comes to speech, dementia patients often face an inability to recall or find the right words for objects and have difficulty regulating their volume and understanding what is said to them.

As more research has come out, we’ve realized that these symptoms manifest sooner and in much subtler ways than we expected. Dementia onset, for example, begins with mild forgetfulness and longer pauses in speech — things that to the physician and average listener would appear as no different than issues caused by ageing. It usually takes far more symptoms through moderate cognitive decline before physicians can make an official diagnosis and administer proper treatment.

But a new type of technology is seeking to change that. With the advent of artificial intelligence and natural language processing, Winterlight Labs have been working to build tools that would allow them to “track, screen for, and predict the onset of these diseases more carefully” [excerpt from Winterlight Labs]. This company uses several resources in linguistics, neuroscience, and machine learning to build and develop a tool that uses artificial intelligence to monitor the speech patterns of their patients and look for those subtle signs of cognitive decline that might otherwise go undetected.

The technology works by analysing hundreds of language snippets of speech and looking for digital biomarkers within those samples. These biomarkers would be consistent among all patients in various stages of dementia. The AI, in its training and testing phases, may even find markers that the programmers did not provide but those that it discovered through its hours of analysis. Once the AI model can accurately identify these digital biomarkers, it can better detect the onset of dementia far earlier than trained physicians could.

Winterlight doesn’t limit their work to just dementia. Since developing their technology, their reach has expanded to other illnesses that display speech-related symptoms such as Alzheimer’s, early signs of depression, multiple sclerosis, schizophrenia, and more.

Each of these illnesses has different causes and symptoms, but they all have the common component of speech impairment. Alzheimer’s, much like dementia, causes individuals to have trouble recalling words or finding the correct terms for objects, but those suffering from Alzheimer’s may also stutter more and find it harder to finish sentences. With depression, the changes are far more subtle. Instead of one’s vocabulary being affected, it is their tone and speed that change. Individuals who are depressed have been shown to have lower, more monotone and laboured tones and speak more slowly with more starts and stops.

Multiple sclerosis has yet another quality of voice that changes: the patient’s harshness and volume. Those suffering from MS are often louder than usual and speak more harshly, enunciating and articulating phrases that a healthy individual might not. Schizophrenic patients, meanwhile, have difficulty concentrating due to the condition, and this manifests in their speech. Those with schizophrenia might speak incoherently, respond to questions with entirely unrelated answers and shift topics rapidly.

While all these symptoms seem different, the AI models Winterlight has developed would be able to adapt to these changes with enough samples.

Artificial intelligence is slowly showing itself as the future of diagnosing illnesses. From cancer to neurological conditions and infections, artificial intelligence models are consistently outperforming expert physicians in the fields, and this trend will only grow as computer power increases. And as diagnosis gets better, medicine and treatments will grow with it too, allowing for faster, more efficient treatment of diseases that, in the past, would have gone unnoticed.

We are entering a new age of healthcare, and it is developing rapidly.

This article was originally published on Medium.