Purpose: To demonstrate that plan-specific optimization of IVD detector position increases the specificity and sensitivity of error detection.
Methods: 19 HDR prostate plans were used to generate virtual simulation of treatment delivery with virtual errors introduced in the plan and virtual IVD monitoring using 4 detectors with realistic characteristics (noise, etc). An algorithm was created to optimize detector position and signal threshold based on plan. Three schemes were investigated: using the algorithm (optimized), using a standard positioning [unoptimized], and positioned at random [random]. A comparison of theoretical use of different detectors (Plastic scintillators, MOSFET, and semiconductor diodes) was performed.
Results: For plastic scintillator detectors, sensitivity and specificity for the optimized scheme(average± SD) were (0.88 ± 0.06), and (0.98 ± 0.01) respectively, in comparison with (0.18 ± 0.06, p <-0.01, T-test), and (0.97 ± 0.07, p >0.05 ) for the unoptimized and ((0.39 ± 0.00, p <0.01, T-test), and (0.94 ± 0.01, p >0.05 )for the random.For the optimized scheme, compared to plastic scintillator detectors, sensitivity (specificity) were 0.88 ± 0.06 p>0.05 (0.98 ± 0.01 p <0.01), 0.88 ± 0.07 p>0.05 (0.95 ± 0.02, p < 0.01), and 0.87 ± 0.08 p > 0.05(0.86 ± 0.06, p < 0.01) respectively for plastic scintillators, semiconductor diodes, and mosfet.
Conclusion: We demonstrated that our algorithm improves sensitivity and specificity using plan-specific IVD optimization when compared with unoptimized and random groups. Permitting the tailoring of IVD at the end of the planning process before treatment, minimizing disruption from false positives and maintaining an acceptable true positive detection rate. Our simulations show that high sensitivity and specificity can be achieved with a range of detector types, although practical and logistical considerations were not considered in this analysis.
Not Applicable / None Entered.