Sarcopenia risk prediction and feature selection by using quantum machine learning algorithms

Sarcopenia is a condition where older individuals experience gradual loss of muscle mass and function. It can be caused by reduced physical activity, hormonal changes, and changes in nutrient intake. Studies have found that sarcopenia increases the risk of falls, fractures, physical disability, hospitalization, and mortality because strong muscles are important for balance, mobility, and overall health. To predict the risk of sarcopenia, this paper explores the use of two quantum models: quantum K-nearest neighbor (QKNN) and amplitude encoding variational quantum classifier (AE-VQC). The models were tested using three sets of experiments with 8, 16, and 32 features, and a feature selection mechanism was used to determine the most important set of features. Both models used an amplitude encoding technique which converts the input data into quantum state amplitude. The QKNN uses the quantum K-minimum-finding algorithm to identify the K closest neighbors of the test state, where the distance among the two quantum states is measured by using fidelity. A swap test is used to produce the statistical assessment of fidelity among the two arbitrary qubits. The VQC model contains a feature map, a variational quantum circuit, a measurement circuit, and a COBYLA optimizer. The models have been analyzed for the desired set of features, and the results are compared with both their classical version and the previously published quantum models. Both the proposed models perform well interms of better accuracy and complexity reduction. The QKNN model was found to be 3%, 3%, and 0.5% more accurate than the classical model for all sets of features in the optimal scenario. Similarly, the AE-VQC model was also 0.3% and 0.5% more accurate than their corresponding classical models for the desired sets of features in the classification of sarcopenia disease.