Title : Unveiling signs of alzheimer through handwriting analysis
Abstract:
Neurodegenerative diseases, such as Alzheimer's and Parkinson's, are distinguished by the progressive loss of motor capabilities, cognitive abilities, or both. The specific origins of neurodegenerative disorders are often unclear; however, they can include genetic, environmental, and lifestyle factors, and are scientifically interpreted as a breakdown of brain and spinal cord functioning [1]. Treatment often focuses on symptom management, as there is presently no cure for these disorders, only treatments to delay disease progression and improve patients' quality of life. Alzheimer's (AD) and Parkinson's (PD) are common neurodegenerative diseases (NDs) that arise for unknown reasons and advance slowly and uncompromisingly. Age that cannot be reversed is one of the most important risk factors for these conditions. Motor deficiencies in PD are driven due to a decline in the basal ganglia's dopaminergic nigrostriatal neurons. On the other hand, AD can be classified as short-term memory loss in its early stages, but as the illness worsens, cognitive and behavioural abilities decline [2]. These critical signs make diseases increasingly serious health and social problems. Although there is no known treatment for brain degeneration, the slow deterioration can be controlled while the disease is still in its early stages. Studies reveal that they significantly impact millions and will sharply rise in the coming decades. Due to the increased risk of dementia, particularly the Alzheimer's type, among patients with mild cognitive impairment (MCI), enhancing the approaches currently used for detecting MCI symptoms is essential. The traditional clinical techniques for detecting degenerative disorders include lumbar puncture, blood testing, and imaging. Additionally, the patient's medical history is beneficial in the early detection of such diseases. Unfortunately, the traditional approaches have many drawbacks, including requiring ongoing sign monitoring and months to obtain a precise diagnosis. The technician's experience is another important factor in the evaluation procedure. For many patients, receiving the proper assistance and care is challenging due to the low diagnosis coverage. Hence, an early diagnosis is crucial to initiate treatment promptly, as significant and irreversible damage may already have occurred once symptoms manifest. One of the first skills affected by cognitive issues is handwriting, which is based on a combination of kinaesthetic and motor perceptual abilities. Therefore, many researchers studied the possibility of using a patient's handwriting to detect early Alzheimer's disease symptoms automatically as it has numerous advantages, including being inexpensive, non-invasive, and non-intrusive. Significant progress has been achieved in this field, beginning with establishing various handwriting protocols that outline the specific writing or drawing tests to be conducted. Nevertheless, it's crucial to emphasise that opinions on the number and kind of tasks that should be used vary widely. However, the majority of studies concurred that some aspects of handwriting, such as writing curves, pen pressure, and speed, can assist in identifying Alzheimer's disease in its early stages. Additionally, there is a scarcity of standardized databases that gather this kind of data, which typically only refers to a relatively small number of participants. This aspect adds another layer of complexity in the realm of machine learning techniques, which typically demand substantial volumes of data. In addition, there is a lack of consensus on the specific characteristics that researchers should prioritize. Indeed, the challenge of identifying effective features that enable the system to differentiate between regular age-related handwriting changes and those induced by neurodegenerative disorders remains unresolved. Finally, researchers must focus on health monitoring related to behavioural traits, such as speech, eye movements, and handwriting, as it will significantly enhance diagnosis coverage and aid in early AD detection. Many scientists are working to find new ways to identify Alzheimer's from handwriting. Here are some current findings in this area:
Azzali I. et al. [3] present a unique method for utilising Vectorial Genetic Programming (VE_GP) to analyse handwriting in order to diagnose Alzheimer's disease (AD). Comprehensive and automated feature extraction is made possible by VE_GP, which analyses raw time series handwriting data without the need for predetermined feature sets. Using a graphic tablet, participants' handwriting was recorded for the study, and pressure and X and Y coordinates were examined. It was discovered that VE_GP achieved better classification performance than both deep learning and conventional feature engineering approaches. It produced models that were easier to understand and more successful in differentiating handwriting patterns linked to AD. The findings illustrated VE_GP's significant advantages over current techniques and showed its potential as a reliable and understandable tool for early AD diagnosis.
Cilia N. et al. [4] investigate a novel method for early diagnosis of Alzheimer's Disease (AD) by analyzing handwriting. The study highlights that handwriting is often one of the first skills affected by AD. To improve diagnostic accuracy, the researchers combined shape and dynamic features from handwriting samples. They generated synthetic color images from online handwriting data, encoding dynamic information such as velocity, jerk, and pressure into the RGB channels. A Convolutional Neural Network (CNN) was then employed to automatically extract features from these images. The study demonstrated that the inclusion of dynamic information significantly enhanced the performance of the diagnostic system compared to using binary images that only contained shape information. The findings indicate that this deep transfer learning approach, which leverages both shape and dynamic features, offers a promising tool for early detection of Alzheimer's Disease. The study found that combining shape and dynamic features in handwriting analysis significantly improves the early diagnosis of Alzheimer's Disease (AD). By creating synthetic color images from online handwriting data, encoding dynamic information such as velocity, jerk, and pressure into the RGB channels, and employing a Convolutional Neural Network (CNN) to extract features, the researchers demonstrated that dynamic features are crucial in distinguishing AD patients from healthy controls. The synthetic images containing both shape and dynamic information outperformed binary images with only shape information, highlighting the potential of this deep transfer learning approach as an effective tool for early AD detection.
Cilia N. et al. [5] investigate the influence of word semantics and phonology on the handwriting of Alzheimer's patients. Using data from six handwriting tasks involving regular words, non-regular words, and non-words, the study employs four machine learning classifiers to analyze kinematic properties of handwriting. The results reveal that non-regular words require more features for effective classification, achieving the highest accuracy of nearly 90%. Feature selection techniques further improved classification performance, with the study demonstrating that handwriting analysis, particularly of non-regular words, can serve as a valuable tool in diagnosing Alzheimer's disease.
Miltra and Rahman [6] employ ensemble machine learning techniques to analyze handwriting kinetics for early diagnosis of Alzheimer's. Utilizing the DARWIN dataset, which includes handwriting samples from 174 individuals, the researchers develop a stacking ensemble model integrating multiple base-level classifiers. The model achieves remarkable performance metrics: 97.14% accuracy, 95% sensitivity, 100% specificity, 100% precision, 97.44% F1-score, 94.37% Matthews Correlation Coefficient (MCC), 94.21% Cohen Kappa, and 97.5% AUC-ROC. This study underscores the potential of machine learning in providing highly accurate and non-invasive diagnostic tools for Alzheimer's disease through handwriting analysis.
In conclusion, machine learning-powered handwriting analysis presents a viable non-invasive and affordable approach to Alzheimer's disease early diagnosis. Important insights into neurodegenerative alterations can be gained by studying the fine motor control needed for handwriting. High accuracy and reliability have been shown in differentiating between healthy persons and those suffering from Alzheimer's disease using advanced machine learning models, such as ensemble classifiers. This method may enhance the quality of life for those who are impacted by it by facilitating early diagnosis and prompt care. The effectiveness of these methods indicates how well they might be incorporated into clinical settings, becoming a useful instrument in the battle against Alzheimer's.
Audience Take Away Notes:
- The cause of neurodegenerative disease
- The risk factors of Alzheimer disease
- The current treatment of such disease
- The drawbacks of current treatment of such disease
- The important of handwriting analysis to help in revealing the early symptoms of Alzheimer
- The most important handwriting features used to detect Alzheimer
