In a recent study, researchers evaluated a fall-risk classification system for people with lower-limb amputations that used artificial intelligence (AI) for foot-strike detection. While an AI step-based feature calculation can be used for general fall-risk analysis, the variability and instability of walking for people with lower-limb amputations make using traditional AI step detection methods ineffective, according to the study. With that in mind, the researchers added manual labeling for more accurate classification. However, since manual foot strike labelling is time-consuming and not viable for clinical use, the researchers developed a random forest machine learning model to use the manually labeled data. Random forests are machine learning algorithms that consist of multiple decision tree classifiers.
Eighty participants (27 fallers, 53 non-fallers) with lower-limb amputations completed a six-minute walk test (6MWT) with a smartphone at the posterior pelvis. The participants (78 percent men) had transtibial (90 percent), transfemoral (3.8 percent), or bilateral (6.2 percent) amputations. Their mean age was 64.2 years, with ages ranging between 19 and 90 years.
Using the machine learning algorithms, the research team found that automated foot strikes correctly classified 72.5 percent of participants’ fall risk, and manually labelled foot strikes correctly classified fall risk for 80 percent of the participants. Automated foot strike detection would improve the feasibility of implementing a fall-risk classification model in a clinical setting; rule-based algorithms can identify steps in elderly individuals with very high accuracy.
The smartphone signals were collected with The Ottawa Hospital Rehabilitation Centre Walk Test app. Automated foot strike detection was completed using a novel Long Short-Term Memory approach. Step-based features were calculated using manually labelled or automated foot strikes.
Foot strikes from turns during the 6MWT were manually identified and used to calculate features from the smartphone signals to train the random forest model for fall risk classification (i.e., features calculated for each stride).
Manually labelled foot strikes correctly classified fall risk for 64 of 80 participants (accuracy 80 percent, sensitivity 55.6 percent, specificity 92.5 percent). Automated foot strikes correctly classified 58 of 80 participants (accuracy 72.5 percent, sensitivity 55.6 percent, specificity 81.1 percent). Both approaches had equivalent fall risk classification results, but automated foot strikes had six more false positives.
According to the study’s authors, the research demonstrated that automated foot strikes from a 6MWT can be used to calculate step-based features for fall-risk classification in people with lower-limb amputations, and automated foot-strike detection and fall-risk classification could be integrated into a smartphone app to provide clinical assessment immediately after a 6MWT.
The open-access study, “Automated step detection with 6-minute walk test smartphone sensors signals for fall risk classification in lower limb amputees,” was published in PLOS Digital Health.