Modelling and identification of characteristic kinematic features preceding freezing of gait with convolutional neural networks and layer-wise relevance propagation

Table 3 Results of the convolutional neural network (CNN) and support vector machine with linear kernel (LSVC)

All scores are given in terms of accuracy (%), assessing the performance of the DL models (and LSVC) on the fourteen freezers individually (Subject 1–14), with a summarized score for the 2421 and 2258 strides extracted from the fourteen non-freezers and fourteen healthy controls, respectively. For the fourteen freezers, the performance is additionally assessed in terms of the sensitivity (%), specificity (%), positive predictive value (PPV) (%), and negative predictive value (NPV) (%). The asterisk (*) is used to denote the seven freezers that froze during the protocol. The dagger (\(\dagger\)) is used to denote the three freezers that froze off camera. The rounded brackets denote the number of extracted strides. For the fourteen freezers, the number of extracted FGCs were controlled for protocol and class imbalance, as explained in the procedure

ISSN: 1472-6947