Fig. 5

Validation of the model with real patient follow-up data. 6 Experienced readers measured volumetric tumour change involving 10 patients and 71 time points. A two-way evaluation was performed in considering either baseline or nadir measurement as reference to balance the assessments between 68 decreasing and 67 increasing changes. Horizontal axis reports the average response assessed by all readers. Vertical axis reports change assessment for each reader, each of them identified by a spot of different colour. Red lines represent the confidence interval predicted by Geary-Hinckley model. 97.1 % of readers’ assessment fall into predicted range of error. Left to -35 % dotted blue line are represented responding assessments, right to +55 % dotted blue line are represented progressive assessment