Machine Learning-Based Gait Analysis for Assessing Balance Ability in Cancer Patients

During the treatment process, cancer patients often experience a decline in balance ability due to various reasons such as surgery, which increases the risk of falls and poses a serious threat to their health. Therefore, it is crucial to quickly and accurately assess the balance ability of cancer patients to take preventive measures against falls at an early stage. This study aims to establish an objective classification model for the quantitative assessment of balance ability in cancer patients using machine learning-based gait analysis techniques. Using wearable devices such as pressure insoles and IMUs, we collected and organized a gait dataset from 78 patients with varying balance abilities. Through feature engineering and machine learning techniques, we employed various feature selection algorithms to obtain the optimal features and then used these features to evaluate the performance of three different models in assessing patients' balance abilities. Experimental results show that SVM achieved the best performance, with average accuracy, precision, recall, F1 score, and area under the curve (AUC) of the test set being 0.8831, 0.8810, 0.8645, 0.8667, and 0.97, respectively. These results indicate that the proposed gait analysis method has significant practical value in assessing the balance ability of cancer patients.