The Clock Test is a well-known test used in daily clinical neuropsychological practice. We present the results of some works developed in the field of cognitive impairment diagnosis using Machine Learning and Artificial Intelligence techniques.

Machine Learning aims to develop techniques that allow computers to learn, through the applied knowledge of Computer Science and Artificial Intelligence.

To this end, they create programs capable of generalizing behaviors from information provided through examples.

The development of these programs has extended to a wide variety of fields; from information search engines to medical diagnostics, including other applications (e.g., STOCK MARKET: a good part of the purchase orders are made by algorithms based on BIG DATA; JUDICIAL JUDGMENTS, etc.).

Within medical diagnostics, we are particularly interested in the development carried out in the field of dementia and cognitive impairment in general.

So far, the clinical diagnosis has been made by the judgment of 1 or more professionals, based on the results of specific tests and reports.

Machine Learning develops algorithms, through an automatic process, from neuropsychological and demographic data.

The objective is to predict outcome in cognitive impairment scales (e.g., Hughes CDR) and clinical diagnoses (e.g., Alzheimer’s disease -EA-, Parkinson’s disease -EP-, etc.).

A large number of algorithms have been used, but 4 of them have provided the best results, the best being BAYESIAN NETWORKS (80% diagnostic accuracy).

The CLOCK TEST is a widely used test in neuropsychological assessment:

• It allows or facilitates the differential diagnosis between normal aging and dementia.
• Minimum economic cost, simple and effective.
• It consists of drawing on a piece of paper the face of a clock whose hands point to 11:10.

In addition, the drawing to the copy is requested. The copy drawing can facilitate the differential diagnosis between different dementias (e.g., the copy drawing in Alzheimer’s disease is better than in dementia with Lewy bodies -DCLw-).

There is a wide variety of classic scoring systems (only with pencil and paper, and the judgment of an expert evaluator). We have been using the scales of the 7 Minute Test.

But recently, alternative scoring systems have appeared, based on Machine Learning.

A group from the Computer Science and Artificial Intelligence Laboratory at MIT (Massachusetts Institute of Technology) has been actively involved in the development of these techniques.

Of these scoring systems, the one developed by Bennasar, 2015 stands out.

• For daily clinical practice it seems very complex because of the large number of characteristics to be analyzed, and because of the exhaustive analysis of the different components of the drawing.
• The scoring system proposed by the MIT group is more in line with the requirements of daily clinical practice.
• It is performed by recording the drawing from an electronic pen that is connected to a computer.
• A computer program developed specifically for this purpose analyzes it.
• The analysis takes into account everything from run time (with a time resolution of 12 ms) to purely graphical data (resolution of ± 0.05 mm: hands, numbers, dial, etc.).

They have used different algorithms and have also created a scoring system without using such a complex analysis. The results are superior to those obtained by the best classical scoring systems.

The scoring system is shown in the following table, and they conclude that when the score obtained is less than 10, it is suggestive of cognitive impairment. With this system they correctly classify about 80% of the cases.

Some interesting results have been reported in papers published so far:

• There is a difference between people with AD/Mild Cognitive Impairment -MCI-/SANOS by analyzing the time it takes to draw the first hand and the total time it takes to draw the clock.
• The total time the patient is drawing, excluding latency time, is longer in patients with Vascular Dementia compared to other dementias and MCI groups. This may be a biomarker for the assessment of bradypsychia, and may aid in the identification of comorbid vascular disease in dementia and MCI.
• The presence of unidentifiable elements may represent difficulties in decision making in timed tasks. This has not been studied so far and may be indicative of impairment even in the context of an apparently correct drawing.

In this way they have been able to correctly identify between 96-99% of the drawings.

The use of the Clock Test is being extended to other pathologies (Multiple Sclerosis, Depression).

BIBLIOGRAPHY .

• Clinical Decision Support System for Early Detection and Diagnosis of Dementia. Mohamed Bennasar. Cardiff University. School of Engineering. PhD Thesis 2014.

• Learning Classification Models of Cognitive Conditions from Subtle Behaviors in the Digital Clock Drawing Test. William Souillard-Mandar, Randall Davis, Cynthia Rudin, Rhoda Au, David J. Libon, Rodney Swenson, Catherine C. Price, Melissa Lamar, Dana L. Penney, 2015.