Parameter calibration from patient data demonstrates model predictivity.
2016-06-10T06:57:22Z (GMT) by
<p>(<i>A</i>) Nonlinear regression analysis of Eq. S2 (coefficient of determination <i>R</i><sup>2</sup> = 0.86) to the measurements of kill fraction and blood volume fraction BVF from histopathology images of 21 patients with CRC metastatic to liver (standard deviations reflect variability of measured values across 20 slides per patient). Inset: parameter values obtained from fit. (<i>B</i>) Linear regression analysis of Hounsfield Unit measurements from pre-treatment arterial-phase contrast-enhanced CT data from 18 patients and blood volume fraction (BVF) measurements from histopathology leads to calibration of BVF parameter (inset). (<i>C</i>) Side-by-side boxplots of <i>f</i><sub>kill</sub> values measured from histopathology and predicted by mathematical model Eq. S2 based on calibration in <i>A</i> and <i>B</i> (18 data points in each set, symbols). In each boxplot, the thick horizontal line is the median; the box is defined by the 25th and 75th percentiles (lower and upper quartile); the diamond is the mean. A paired t-test at the 0.05 significance level resulted in <i>P</i> = 0.44, indicating that the observed difference between the two data sets is not significant. (<i>D</i>) Predictions of Eq. S2 (open circles, average relative error ≈ 24%) compared, for each patient, to the direct measurements from histopathology post-treatment and resection (filled circles, with standard deviation of multiple measurements per patient).</p>