Application of machine learning techniques for dementia severity prediction from psychometric tests in the elderly population

Previous research has shown the benefits of early detection and treatment of dementia. This detection is usually performed manually by one or more clinicians based on reports and psychometric testing. Machine learning algorithms provide an alternative method of prediction that may contribute, with an automated process and insights, to the diagnosis and classification of the severity level of dementia. The aim of this study is to explore the use of neuropsychological data from a reduced version of the Addenbrooke's Cognitive Examination III (ACE-III) to predict absence or different levels of dementia severity using the Global Deterioration Scale (GDS) scores through the implementation of the kNN machine learning algorithm. A sample of 1164 elderly people over sixty years old were evaluated using a reduced version of the ACE-III and the GDS. The kNN classifier provided good accuracies using 15 items from the ACE-III and adequately differentiating people with absence and mild impairment, from those with more severe levels of impairment according to the GDS rating. Our results suggest that the kNN algorithm may be used to automate aspects of clinical cognitive impairment classification in the elderly population.

Psychometric properties of Addenbrooke's Cognitive Examination III (ACE-III): An item response theory approach

The Addenbrooke's Cognitive Examination III is one of the most widely used tests to assess cognitive impairment. Although previous studies have shown adequate levels of diagnostic utility to detect severe impairment, it has not shown sensitivity to detect mild decline. The aim of this study was to evaluate the psychometric properties of Addenbrooke's Cognitive Examination III in a large sample of elderly people through Item Response Theory, due to the lack of studies using this approach. A cross-sectional study was conducted with 1164 people from the age of 60 upwards, of which 63 had a prior diagnosis of Alzheimer dementia. The results showed that, globally, the Addenbrooke's Cognitive Examination III possesses adequate psychometrics properties. Furthermore, the information function test shows that the subscales have different sensitivity to different levels of impairment. These results can contribute to determining patterns of cognitive deterioration for the adequate detection of different levels of dementia. An optimized version is suggested which may be an economic alternative in the applied field.

Revisiting linear discriminant techniques in gender recognition

Emerging applications of computer vision and pattern recognition in mobile devices and networked computing require the development of resource-limited algorithms. Linear classification techniques have an important role to play in this context, given their simplicity and low computational requirements. The paper reviews the state-of-the-art in gender classification, giving special attention to linear techniques and their relations. It discusses why linear techniques are not achieving competitive results and shows how to obtain state-of-the-art performances. Our work confirms previous results reporting very close classification accuracies for Support Vector Machines (SVMs) and boosting algorithms on single-database experiments. We have proven that Linear Discriminant Analysis on a linearly selected set of features also achieves similar accuracies. We perform cross-database experiments and prove that single database experiments were optimistically biased. If enough training data and computational resources are available, SVM's gender classifiers are superior to the rest. When computational resources are scarce but there is enough data, boosting or linear approaches are adequate. Finally, if training data and computational resources are very scarce, then the linear approach is the best choice.