Integrating Multi-sensor Time-series Data for ALSFRS-R Clinical Scale Predictions in an ALS Patient Case Study.

Area of Science:

• Neurology
• Biomedical Engineering
• Data Science

Background:

• Current amyotrophic lateral sclerosis (ALS) monitoring relies on infrequent, in-clinic assessments like the ALS Functional Rating Scale Revised (ALSFRS-R).
• This limited data collection can delay crucial treatment adjustments by missing subtle, day-to-day changes in patient functional status.
• Objective, continuous monitoring is needed to capture early indicators of disease progression.

Purpose of the Study:

• To explore the use of in-home health sensors for frequent, quantitative tracking of functional status in ALS patients.
• To develop and evaluate methods for integrating high-frequency sensor data with lower-frequency clinical assessments (ALSFRS-R).
• To assess the potential of sensor-based health metrics as sensitive prognostic markers for ALS.

Main Methods:

• Utilized the XGBoost regressor for base learning.
• Investigated nine different interpolation models to align monthly ALSFRS-R targets with high-frequency sensor-based health features.
• Employed a mixed-frequency data modeling approach.

Main Results:

• The evaluated interpolation models demonstrated superior prediction of ALSFRS-R scores compared to traditional linear slope estimates.
• Sensor-based health estimates showed potential as sensitive prognostic markers.
• The study successfully modeled mixed-frequency data from clinical assessments and home sensors.

Conclusions:

• In-home health sensors offer a practical supplement to traditional clinical tools for monitoring ALS progression.
• Sensor-derived data can provide more frequent and quantitative insights into patient functional status.
• This approach holds promise for earlier detection of health shifts and timely intervention in ALS management.