Extended Data Fig. 5: Comparison of results from linear support vector machine (SVM), linear regression, and neural network (NN) regression.
From: A swallowable X-ray dosimeter for the real-time monitoring of radiotherapy
a, Measured radioluminescence intensity over time at a dose of 10 mGy and dose rates from 4 to 16 mGy/min. b and c, Comparison of the accuracy of dose estimation using different algorithms and feature parameter selection approaches, respectively. d–f, Regression results using linear SVM (d), linear regression (e), and the NN regression algorithm (f), with prediction-true value scatter plots (top panels), residual plots (middle panels) and RMSE under different feature strategies (bottom panels). When radioluminescence intensity (L), afterglow intensity (A) at different times, and temperature (T) are selected as feature parameters of the regression algorithms, the RMSE of the three algorithms is 0.304, 0.141, and 0.04 mGy/min, respectively. The RMSE statistics for each algorithm are derived from 200 executions. In c–f, bar plots and error bars show the mean ± s.d.
