A research group used artificial intelligence (AI) to process natural language and assess speech characteristics of Parkinson’s disease (PD) patients. His AI analysis of the data found that these patients spoke using more verbs and fewer nouns and fillers. The study was led by Professor Masahisa Katsuno and Dr. Katsunori Yokoi of Nagoya University Graduate School of Medicine, in collaboration with Aichi Prefectural University and Toyohashi University of Technology. They published their results in the journal Parkinson’s Disease and Related Disorders.
Natural language processing (NLP) technology is a branch of AI focused on enabling computers to understand and interpret large amounts of human language data using statistical models that identify patterns. Considering that PD patients experience a variety of speech-related problems, such as impaired speech production and language use, the group used NLP to develop 37 features using texts created from free speech. We analyzed the differences in patients’ voice patterns based on
Analysis revealed that PD patients used fewer common nouns, proper nouns, and fillers per sentence. On the one hand, they spoke using more verb-per-sentence ratios and case particle variations (an important feature of Japanese).
According to Yokoi, “When asked to talk about their day in the morning, a PD patient might say something like, ‘I woke up at 4:50 in the morning.’ I thought maybe, but I woke up.It took about 30 minutes to go to the toilet, so I washed my face and changed clothes around 5:30 in the morning.My husband made breakfast.After 6:00 in the morning. Then I brushed my teeth and got ready to go.”
Yokoi continued. “A healthy control group might say something like, ‘In the morning I got up at 6 o’clock, got dressed, yes, I washed my face. Then I fed my cat and dog. My daughter prepared food, but I said I couldn’t eat it, and I drank water.”
“These are examples of conversations created to reflect the characteristics of PD patients and healthy people, but what I want you to see is that the overall length is similar,” explains Yokoi. “However, because PD patients speak shorter sentences than people in the control group, more verbs are used in the machine learning analysis. Fillers such as “hmm” are used more often. “
The most promising aspect of this study is that the team tested patients who had not yet exhibited the characteristic cognitive decline seen in PD. Their findings therefore provide a potential means of early detection to distinguish PD patients.
“Our results suggest that even in the absence of cognitive decline, the conversations of PD patients differ from those of healthy subjects,” concluded study director Professor Katsuno. “Based on these speech changes, he tried to identify a PD patient or a healthy control and was able to identify her PD patient with over 80% accuracy. It suggests the possibility of diagnosing PD by linguistic analysis.”
