Eyetracking metrics reveal impaired spatial anticipation in behavioural variant frontotemporal dementia

Silvia Primativo, Camilla N. Clark, Keir Yong, Nicholas C. Firth, Jennifer M. Nicholas, Daniel C. Alexander, Jason D. Warren, Jonathan D. Rohrer, Sebastian J. CrutchNeuropsychologia 2017 · DOI

TL;DR

This study introduces an innovative, instruction-less eyetracking task to assess spatial anticipation, a complex cognitive function often impaired in frontotemporal dementia (bvFTD). By analyzing eye movement patterns, the researchers found that bvFTD patients exhibit significant deficits in predicting upcoming visual targets, a finding more sensitive than traditional pen-and-paper tests. Machine learning further enhanced the ability to detect these subtle impairments, suggesting eyetracking as a powerful tool for early diagnosis and monitoring of neurodegenerative diseases.

What Problem Does This Paper Address?

Diagnosing and understanding the progression of neurodegenerative diseases like frontotemporal dementia (FTD) can be challenging. While basic eye movement abnormalities are sometimes observed, they often don't capture the full spectrum of cognitive decline, particularly higher-order executive functions. Traditional cognitive tests, while valuable, can be confounded by linguistic demands, patient fatigue, and the need for complex instructions, potentially masking subtle deficits. Specifically, the behavioral variant of FTD (bvFTD) is characterized by insidious changes in personality, behavior, and executive functions, making early and accurate diagnosis difficult. The Brixton spatial anticipation test is known to be sensitive to frontal lobe damage, but its reliance on verbal instructions can be a barrier for some patients. This paper aims to address this by developing and validating an eyetracking-based adaptation of the Brixton test that minimizes linguistic input and leverages the rich data from eye movements to reveal more nuanced cognitive impairments.

Methods

The core of this research lies in the development of a novel eyetracking paradigm and its analysis using both frequentist and Bayesian machine learning approaches.

The Eyetracking Task: The researchers adapted the Brixton spatial anticipation test into a computerized, instruction-less format. Participants were presented with a 10x7 matrix of circles. In each trial, a black dot moved sequentially through seven positions, following one of 12 predefined patterns (straight line, zigzag, displaced zigzag). The participant's only task was to press a button whenever they saw the target move to a new position. This design aimed to minimize reliance on verbal instructions and focus solely on the visual-motor response to predict the target's movement.

Data Analysis: 1. Frequentist Statistical Analysis: Standard eye movement metrics (e.g., saccade latency, fixation duration) were analyzed. Crucially, the researchers also quantified "anticipatory saccades" – eye movements made towards the predicted location of the next target before it appeared. These were classified as "correct" (if they accurately predicted the next target) or "incorrect" (if they did not). 2. Bayesian Machine Learning (ML) Approach: A more advanced analysis involved using raw eyetracking time-series data. A Bayesian ridge regression model was trained on data from healthy controls to predict a subject's next gaze location based on their previous movements. This model was then applied to data from bvFTD and semantic dementia (SD) patients to assess how well it could predict their gaze patterns, effectively identifying deviations from typical control behavior.

Datasets

The study utilized data from three groups:

  • Behavioral Variant Frontotemporal Dementia (bvFTD) patients: N=12
  • Semantic Dementia (SD) patients: N=6
  • Healthy Controls: N=38

The eyetracking data was collected during the experimental task. For the machine learning approach, the control group data was split into training, validation, and test sets, while bvFTD and SD patient data were used solely for testing. No new benchmark datasets were introduced; rather, the existing Brixton spatial anticipation test served as a point of comparison.

Key Insights and Ideas

  • Eyetracking as a Sensitive Probe for Executive Function: The study demonstrates that detailed analysis of eye movements, beyond basic oculomotor functions, can reveal subtle deficits in higher-order cognitive processes like spatial anticipation, even when traditional tests are less conclusive.
  • Instruction-less Design Minimizes Confounds: Removing explicit verbal instructions significantly reduced the influence of language impairments and potentially other cognitive confounds, allowing for a purer measure of spatial anticipation.
  • bvFTD Patients Show Distinct Anticipatory Saccade Patterns: The key finding is that bvFTD patients make significantly more incorrect anticipatory saccades and fewer correct ones, particularly in more complex zigzag patterns, compared to controls and SD patients. This suggests a breakdown in predictive processing and/or inhibition of erroneous predictions.
  • Machine Learning Enhances Diagnostic Power: The Bayesian ML approach proved more sensitive than frequentist methods, identifying significant differences between bvFTD patients and controls across all individual trials, and between SD patients and controls in a substantial subset of trials. This highlights the potential of ML for detecting subtle impairments and optimizing diagnostic tools.
  • Potential for Early Detection and Monitoring: The findings suggest that this eyetracking paradigm could be a valuable tool for early detection of spatial anticipation deficits in at-risk individuals and for monitoring disease progression.

Key Results

  • Traditional Brixton Test: Identified deficits in 58% of bvFTD patients and 33% of SD patients.
  • Frequentist Eyetracking Analysis:
    • Identified deficits in 92% of bvFTD patients.
  • Found no deficits in SD patients, suggesting the original test's sensitivity might be partly due to its instructions.
  • bvFTD patients made significantly more incorrect anticipatory saccades than controls and SD patients, especially in zigzag conditions.
  • Machine Learning Analysis:
    • Identified significant differences between bvFTD and controls in 24/24 trials.
  • Identified significant differences between SD and controls in 12/24 trials.
  • The ML approach was more sensitive than frequentist methods in detecting group differences, particularly for SD patients.
  • Reaction Times: bvFTD patients showed slower button press reaction times, which the authors attribute to the need for corrective saccades following incorrect anticipations, rather than general motor slowing.

Limitations and Future Work

  • Sample Size: The study acknowledges that the sample sizes, particularly for the patient groups (bvFTD: N=12, SD: N=6), are limited. This may have contributed to the null findings in the neuroimaging analysis and suggests the need for larger studies.
  • Interpretability of ML Models: While powerful, the ML approach's predictive error metrics are less directly interpretable than traditional eyetracking measures like fixation duration. The authors suggest combining both approaches to leverage their respective strengths.
  • Generalizability of ML: The ML models were trained on control data. While this is a standard approach for detecting deviations, further research could explore models trained on patient data or more complex cross-validation strategies.
  • Future Directions:
    • Clinical Application: The researchers propose using this paradigm for individuals at risk of FTD and for clinical trials to assess treatment efficacy.
  • Technological Advancement: Adapting the test for cheaper, more accessible eyetracking systems is a key goal.
  • Broader Cognitive Domains: The ML approach's potential for analyzing complex time-series data could be extended to other cognitive domains beyond spatial anticipation.
  • Neuroimaging Integration: Further investigation into the neural correlates of these eyetracking deficits, with larger sample sizes, is warranted.

Summary updated

Abstract

Eyetracking technology has had limited application in the dementia field to date, with most studies attempting to discriminate syndrome subgroups on the basis of basic oculomotor functions rather than higher-order cognitive abilities. Eyetracking-based tasks may also offer opportunities to reduce or ameliorate problems associated with standard paper-and-pencil cognitive tests such as the complexity and linguistic demands of verbal test instructions, and the problems of tiredness and attention associated with lengthy tasks that generate few data points at a slow rate. In the present paper we adapted the Brixton spatial anticipation test to a computerized instruction-less version where oculomotor metrics, rather than overt verbal responses, were taken into account as indicators of high level cognitive functions. Twelve bvFTD (in whom spatial anticipation deficits were expected), six SD patients (in whom deficits were predicted to be less frequent) and 38 healthy controls were presented with a 10 × 7 matrix of white circles. During each trial (N = 24) a black dot moved across seven positions on the screen, following 12 different patterns. Participants' eye movements were recorded. Frequentist statistical analysis of standard eye movement metrics were complemented by a Bayesian machine learning (ML) approach in which raw eyetracking time series datasets were examined to explore the ability to discriminate diagnostic group performance not only on the overall performance but also on individual trials. The original pen and paper Brixton test identified a spatial anticipation deficit in 7/12 (58%) of bvFTD and in 2/6 (33%) of SD patients. The eyetracking frequentist approach reported the deficit in 11/12 (92%) of bvFTD and in none (0%) of the SD patients. The machine learning approach had the main advantage of identifying significant differences from controls in 24/24 individual trials for bvFTD patients and in only 12/24 for SD patients. Results indicate that the fine grained rich datasets obtained from eyetracking metrics can inform us about high level cognitive functions in dementia, such as spatial anticipation. The ML approach can help identify conditions where subtle deficits are present and, potentially, contribute to test optimisation and the reduction of testing times. The absence of instructions also favoured a better distinction between different clinical groups of patients and can help provide valuable disease-specific markers.
BibTeX
@inproceedings{primativo2017eyetracking,
  author = {Silvia Primativo and Camilla N. Clark and Keir Yong and Nicholas C. Firth and Jennifer M. Nicholas and Daniel C. Alexander and Jason D. Warren and Jonathan D. Rohrer and Sebastian J. Crutch},
  title = {{Eyetracking metrics reveal impaired spatial anticipation in behavioural variant frontotemporal dementia}},
  booktitle = {Neuropsychologia},
  year = {2017},
  doi = {10.1016/j.neuropsychologia.2017.10.014},
}

Loading related content...

Figure