Using Explainable Artificial Intelligence in the Clock Drawing Test to Reveal the Cognitive Impairment Pattern.

The prevalence of dementia is currently increasing worldwide. This syndrome produces a deterioration in cognitive function that cannot be reverted. However, an early diagnosis can be crucial for slowing its progress. The Clock Drawing Test (CDT) is a widely used paper-and-pencil test for cognitive assessment in which an individual has to manually draw a clock on a paper. There are a lot of scoring systems for this test and most of them depend on the subjective assessment of the expert. This study proposes a computer-aided diagnosis (CAD) system based on artificial intelligence (AI) methods to analyze the CDT and obtain an automatic diagnosis of cognitive impairment (CI). This system employs a preprocessing pipeline in which the clock is detected, centered and binarized to decrease the computational burden. Then, the resulting image is fed into a Convolutional Neural Network (CNN) to identify the informative patterns within the CDT drawings that are relevant for the assessment of the patient's cognitive status. Performance is evaluated in a real context where patients with CI and controls have been classified by clinical experts in a balanced sample size of [Formula: see text] drawings. The proposed method provides an accuracy of [Formula: see text] in the binary case-control classification task, with an AUC of [Formula: see text]. These results are indeed relevant considering the use of the classic version of the CDT. The large size of the sample suggests that the method proposed has a high reliability to be used in clinical contexts and demonstrates the suitability of CAD systems in the CDT assessment process. Explainable artificial intelligence (XAI) methods are applied to identify the most relevant regions during classification. Finding these patterns is extremely helpful to understand the brain damage caused by CI. A validation method using resubstitution with upper bound correction in a machine learning approach is also discussed.

Hypoalbuminemia and other prognostic factors of mortality at different time points after ischemic stroke.

OBJECTIVE: The aim of this study was to investigate whether hypoalbuminemia and other risk factors for mortality after stroke have the same or different short (1 month), medium (3 months), long (1 year) or very long term (5 years) prognostic value. SUBJECTS/METHODS: clinical and analytical data from 254 patients admitted to our Hospital with an ischemic stroke and followed up prospectively for 2 years were collected with a prospective standard protocol. Additional data up to 5 years were obtained from Clinical and Laboratory Registries of the Hospital, a mailed questionnaire, a phone call and the Council Registry of Mortality. Risk factors for mortality at different time points were calculated with logistic regression and Cox proportional hazard analyses. RESULTS: The following factors were significantly associated with mortality at one month: cardioembolic mechanism, hypoalbuminemia, glycemia, age, low diastolic arterial pressure and Canadian Scale, at three months: previous stroke and Barthel index at discharge, at one year: previous dementia and Barthel index at three months and at five years: age, Canadian Scale score at discharge and low cholesterol at admission. Cox regression analysis considering survival time showed hypoalbuminemia at admission (hazard ratio (HR) 2; p = 0.03), age (HR 1.06; p < 0.00), previous dementia (HR 2; p < 0.00), cardioembolic mechanism (HR 2; p < 0.00) and severity on the Canadian Neurological Stroke Scale (HR 1.2; p < 0.00) to be independently associated with mortality. CONCLUSION: Mortality after ischemic stroke seems to depend on different factors along time. Hypoalbuminemia at admission is an independent factor for short term (acute) and global mortality. Other risk factors for global mortality were previous dementia, cardioembolic mechanism and severity on the Canadian Neurological Stroke Scale at admittance.

Objetivo: El propósito del estudio era investigar si la hipoalbuminemia y otros factores de riesgo de mortalidad tras un ictus tenían el mismo valor pronóstico tras un ictus a corto (1 mes), medio (3 meses), largo (1 AÑO) o muy largo plazo (5 AÑOs). Métodos: Se estudiaron 254 pacientes ingresados en nuestro hospital con ictus isquémico y seguidos prospectivamente durante dos AÑOs con un procolo estándar de forma prospectiva. Se recogieron datos adicionales hasta 5 AÑOs de las Historias Clínicas, los datos del laboratorio, un cuestionario enviado por correo, una llamada telefónica y la revisión de los Registros de Mortalidad de los ayuntamientos. Los factores de riesgo de mortalidad en cada periodo se calcularon con regresión logistica y el modelo de riesgos proporcionales de Cox. Resultados: Se asociaron de forma significativa con la mortalidad al mes el mecanismo cardioembólico, la hipoalbuminemia, la glucemia al ingreso, la edad, la presión arterial diastólica más baja y la puntuación en la Escala Canadiense. A los tres meses, la existencia de ictus previos y el índice de Barthel al alta. Al AÑO la existencia previa de demencia y el índice de Barthel a los 3 meses y a los cinco AÑOs la edad, la puntuación en la escala Canadiense al alta y un colesterol menor al ingreso. El análisis de regresión de Cox considerando el tiempo de supervivencia, mostró una asociación independiente con la mortalidad de la hipoalbuminemia al ingreso ((HR) 2; p = 0,03), la edad (HR 1,06; p < 0,00), la demencia previa (HR 2; p < 0,00), el mecanismo cardioembólico (HR 2; p < 0,00) y la severidad según la escala Canadiense (HR 1.2; p < 0,00). Conclusión: La mortalidad tras un ictus isquémico parece depender de distintos factores según el tiempo transcurrido.

[Cognitive assessment in the elderly. Normative data of a Spanish population sample older than 70 years]. / Evaluación cognitiva del anciano. Datos normativos de una muestra poblacional española de más de 70 años.

BACKGROUND AND OBJECTIVE: To define the parameters of normal cognitive performance in an easy to administer neuropsychological battery of a Spanish population sample of elderly over 70 years. SUBJECTS AND METHOD: We examined 527 home dwelling elderly from the longitudinal study Envejecer en Leganés, 71 to 99 years old (mean age [SD] 79 [6.2] years, 51.8% women, 38.7% without formal education), with an extensive clinical survey, a neurological exam and the following neuropsychological battery: Short Portable Mental Status Questionnaire, Mini Mental State Examination, Benton Orientation Test, Bell Test, Verbal Fluency, Clock Drawing Test, Trail Making Test, Free and Cued Figures Recall, Logic Memory, Naming, incidental Recall, Delayed Recall, Similarities, IQCODE Questionnaire Of Jorm and Depression Questionnaire CES-D. 111 cases who rejected the assessment or had sensomotor limitations were excluded. 368 non demented subjects (according to DSM-IV criteria; 88.5% of the de 416 remaining cases) were selected to establish the normative data of every test. RESULTS: The analyzed subsample was representative of the total sample. The normative data for every neuropsychological test were obtained in the total group of non demented subjects and in the subgroups stratified by four age levels and two cultural levels. The mean values, standard deviations, range and percentiles for every test and subgroup are presented. CONCLUSIONS: The normative data obtained in this population sample for a neuropsychological battery easy to administer to poorly educated elderly can be useful to rigourously apply the criteria of dementia and mild cognitive impairment in future clinical and population studies.