Predicting the lifespan of cancer patients by replicating known quantum embedding processes in novel circuit schemes in Qiskit

The lifespan predictions oncologists make to schedule treatments for oncologists are extremely error-prone; about 74% of patients get a prediction with an error margin of a year, over which the prediction becomes inaccurate after the life expectancy increases over a year. One potential tool proposed to aid oncologists in lifespan prediction is through the creation of a machine learning model; however, classical machine learning cannot process the data well due to the lack of structure/pattern in lifespan data and the lack of public lifespan data available. We propose to use a quantum embedding procedure to predict the life expectancy of a cancer to account for the noisiness and sparsity of clinical data by using various quantum circuits to encode classical data into quantum states. Approximately 5000 points of cancer data from NCI TCGA are encoded into the IQP, angle, and amplitude embeddings to elevate the clinical data into a higher dimensional space known as the Hilbert space, which serves as the input data into an SVR algorithm in the form of a 2D matrix. The metric of accuracy will be the mean absolute error (MAE) between the model’s predictions and the true patient lifespan in years. This method demonstrates that the quantum models have a lower error margin than the classical models, validating the quantum supremacy concept and allowing for accurate lifespan prediction for cancer patients and more optimal treatment plans.