Development of an intelligent system for diagnosing cognitive impairments based on machine learning methods and handwriting motor analysis

The global rise in dementia is becoming one of the key medical and social challenges of the 21st century: according to World Health Organization estimates, the number of patients has already exceeded 55 million and is growing by about 10 million every year. Alzheimer's disease predominates, accounting for 60–70% of cases; it is often preceded clinically by mild cognitive impairment (MCI). The earlier MCI is detected, the greater the chance of slowing or preventing the progression to dementia.

In Kazakhstan, the situation is aggravated by fragmented statistics: expert estimates put the number of people with cognitive disorders at 100–120 thousand, while official figures are considerably lower. The diagnostic gap is caused by three main factors:

• a limited number of specialized neurologists and gerontologists, especially in the regions;
• the high cost of instrumental methods (MRI, CT) and biomarkers;
• the lack of scalable, language-adapted cognitive tests.

Traditional scales (MMSE, MoCA) remain useful but are sensitive to language and educational level, and repeated testing leads to a "learning effect". Invasive or expensive methods (CSF biomarker analysis, neuroimaging) are unsuitable for mass screening. Under these conditions, there is a demand for a non-invasive, affordable technology that can detect early markers of cognitive decline quickly and regardless of region.

The use of a sensor pen that records micro-movements during writing opens up a new direction in diagnostics. Time series of pen coordinates, pressure and acceleration contain hidden information about motor and cognitive functions. Fluctuation analysis methods (DFA, spectral and wavelet approaches) make it possible to extract objective micro-patterns, and machine learning algorithms can translate them into diagnostic conclusions. Such a tool is easy to integrate into general practitioners' offices or home monitoring systems, which will broaden coverage and reduce the workload of specialists. In addition, building our own handwriting motor datasets creates a scientific foundation for further research in neuroinformatics.

Thus, the development of an intelligent system based on handwriting analysis and artificial intelligence methods meets a critical need of Kazakhstan's healthcare system — affordable, scalable and early diagnosis of cognitive impairments that can significantly reduce the socio-economic burden of dementia.