# Impulsivity in cerebellar ataxia: an online, multidimensional assessment

## TL;DR

This study explores how cerebellar ataxia affects different types of impulsivity using questionnaires and machine learning, revealing a unique impulsivity profile in patients.

## Contribution

The study identifies a distinct impulsivity profile in cerebellar ataxia using large cohorts and machine learning classification.

## Key findings

- CA patients showed higher motor and attentional impulsivity but lower non-planning impulsivity.
- CA patients exhibited reduced impulsivity in monetary decision-making tasks.
- Machine learning models accurately classified CA patients based on impulsivity features.

## Abstract

While considered a motor control structure, the cerebellum contributes to non-motor functions, including impulsivity. However, whether it contributes to impulsivity in a domain-specific manner remains unknown. Studies on cerebellar ataxia (CA), a common model for cerebellar dysfunction, typically have small sample sizes, limiting robustness. In a multicenter cross-sectional study, we investigated the cerebellum’s role in various forms of impulsivity by comparing large cohorts of CA to age- and education-matched neurotypical healthy (NH) controls. Additionally, to examine the ability to identify individuals with CA using impulsivity features alone, we developed supervised machine learning (ML) models. In experiment 1 (CA = 140, NH = 136), impulsivity was assessed using the BIS-11 questionnaire. In experiment 2 (CA = 110, NH = 107), performance-based impulsivity was assessed using the MCQ-27, evaluating delay discounting in monetary decision-making. Two ML models—Logistic Regression and Random Forest—were utilized to classify disorder status (CA/NH). The CA group showed higher BIS-11 scores (p = 0.001), indicating higher impulsivity, driven by motor (p < 0.001) and attention (p = 0.002) impulsivity. However, the CA group exhibited lower non-planning impulsivity (p = 0.014). In the MCQ-27, the CA group showed lower k-values (p < 0.005), indicating reduced impulsivity in monetary decisions. Both ML models demonstrated strong classification performance (AUC ≥ 0.85) in independent datasets. This study highlights the cerebellum’s selective role in impulsivity. We found higher motor and attentional impulsivity in CA alongside lower non-planning and decision-making impulsivity. This suggests a unique impulsivity profile in CA that may indicate a compensatory mechanism for future events. ML models demonstrated high classification performance, suggesting impulsivity is a core non-motor feature of CA.