Brain Tumor Prediction with LSTM Method

The second most prevalent illness in the world, brain tumors cause one-sixth of deaths. This paper aims to provide some guidance for people's healthy life by comparing different brain cancer prediction models and to provide a numerical basis for the targeted use of medical financial resources by predicting the incidence and mortality of cancer in the next few years. Primary data on lifestyle habits and cancer incidence from CDC and American Cancer Society, 1990–2017 were used for analysis. To ensure the quality of data sources, this paper first formats conversion and duplicate data elimination, and uses filters the data to obtain the statistical values required for data analysis. Then, the paper selects the cubic spline interpolation technique with good smoothing performance and suitable for the obtained data source to expand the original data and converts the annual data into monthly data. Finally, LSTM and CNN are used to analyze them and then compare their accuracies. The experiment proves that the smallest CNN is the mean square error and mean absolute percentage error of the LSTM model, and R2 (R-square, correlation coefficient) is closest to 1. Therefore, the LSTM model is more suitable for predicting brain tumors.