Early diagnosis of the disease and the recognition of the so called “prodromal syndrome”, a specific group of symptoms that may precede the onset of the illness, may greatly increase the chances of a successful therapy. Currently, no such screening method is available that is capable of detecting this specific and sensitive condition. According to our research, however, this early, slight cognitive disorder has an effect on the speech of the patient, both on the acoustic form and the verbal content.
The screening test
Our objective is to develop – with the help of a sensitive, interactive software – a spontaneous speech induced, computer assisted screening test that is capable of detecting Alzheimer’s dementia early and can be performed in the patient’s home. To this end, we are developing a combined, visual-verbal test procedure that measures certain parameters of linguistic functions and, in both international and domestic comparison, uniquely detects the factors suitable for preclinical diagnosis. Manual indicator retrieval is a complex and costly procedure and requires skilled labor. Our objective is to extract these indicators automatically, with the help of speech recognition technology.
Machine learning in two steps
The software being developed applies machine learning in two steps: in the first step, the speech recognition system measures the indicators from earlier research as accurately as possible (e.g. speech and articulation rate, number and length of filled and silent pauses); in the second step, on the basis of the properties thusly extracted the statistics based machine learning method gives an estimate of whether the speaker has mild cognitive impairment or not. On this basis, the system will make a recommendation for the test subject to see a specialist.