Time-Series & Sensors

Dual-Relationship Modeling of Friedreich Ataxia

Dual-Relationship Modeling of Friedreich’s Ataxia: Integrating Cross-Sectional and Longitudinal Movement Dynamics

A model that gets both today’s severity and the rate of change right, instead of only one snapshot at a time.

The problem

Sensor models match clinical scores well at a single visit but do not capture how one person’s disease progresses over time.

The approach

Training data is structured as paired longitudinal observations, with a custom objective that rewards accurate point-in-time predictions and plausible, age-aware progression.

The outcome

Strong cross-sectional correlation with better longitudinal consistency than regression-only models.

In detail

Developed a machine-learning framework that models two complementary aspects of Friedreich ataxia at once: cross-sectional disease severity at an individual assessment and longitudinal change across repeated assessments. The approach combines sensor-derived movement features from AIM-C, AIM-S and AIM-P with clinical predictors and age-stratified natural-history progression rates. A custom objective encourages accurate point-in-time severity prediction while maintaining clinically plausible progression trajectories, improving longitudinal consistency over conventional regression-only models.

Built with

• Longitudinal Modelling
• Clinical AI
• IMU Analytics

More in Time-Series & Sensors

• Probabilistic Deep Ensembles for FRDA Severity Assessment
• Cross-Device Domain Adaptation for FRDA Assessment
• AIM-M: Multiview IMU Assessment of Friedreich Ataxia

Next project: Adaptive LSTM-RNN Dynamic Equivalents for Power Systems

By Kanishka Ranaweera, PhD, Machine Learning Engineer in Melbourne, Victoria, Australia. hello@kanishkabuilt.me