Fig. 4: Functional evaluation of predicted distributions under aDBS setting.
From: Transformer-based long-term predictor of subthalamic beta activity in Parkinson’s disease
Evaluation of the predictions assessed by monitoring the inaccuracies of LAURA in identifying the 10th and 90th percentiles of each daily distributions on patient NWK1 (see Suppl. Figure 6 for all patients). a Examples of observed (gray) and predicted (green on the left, blue on the right) distributions in the task 2 days→1 day-ahead within both cases of single aDBS setting (left) and multiple aDBS settings (right). For the sake of clarity, the distributions have been presented as min–max normalized occurrences over beta power, by resampling and taking the average of each successive pair of bins. The 10th and 90th percentiles are reported for the observed (in gray) and the predicted (in dark yellow and dark orange, respectively) distributions. b Inaccuracies of LAURA in identifying the 10th (dark) and 90th (light) percentiles for patient NWK1, for both cases of single (left) and multiple (right) aDBS settings. Results support LAURA as a patient-specific system.
