Using statistical measures for automated comparison of in‐beam PET data

Abstract

The results demonstrate that the automated comparison using PCC provides similar results in terms of sensitivity and specificity compared to visual inspections of in-beam PET data. Hence the method presented in this study is a promising and effective approach to improve the efficiency in the clinical workflow in terms of particle therapy monitoring by means of PET.

Clinical implementation and range evaluation of in vivo PET dosimetry for particle irradiation in patients with primary glioma

“…Some research groups in Japan [16,17], the USA [18,19] and at GSI Darmstadt, Germany [12,20] have already gathered some experience concerning the in vivo monitoring and several authors (Frey et al [21], Knopf et al [22], Min et al [23], Helmbrecht et al [24], Kuess et al [25]) have investigated strategies to evaluate the beam range and thereby to validate the accuracy of particle therapy.…”

Systematic analysis on the achievable accuracy of PT-PET through automated evaluation techniques

“…The use of an additional median filter on the PT-PET images prior to comparison showed some promising results during a previous pilot study, where regarding five patient test cases the detection capability of 6 mm range modified data could be improved by 10% by using a median filter, with a kernel size of 11 × 11 × 11 voxels [19]. Therefore, the evaluation of phantom A was also used to evaluate quantitative differences regarding the PCC evaluation between images that were pre-filtered and images that were not filtered before evaluation.…”

“…The second investigated automatic evaluation method utilizes the Pearson correlation coefficient (PCC) as outlined in Kuess et al [19,21] to compare the reference β + -activity distribution (no range modification or hollow cavities) to the respective β + -measurement. The PCC, as depicted in Eq.…”

“…Furthermore, this task adds a massive workload to the clinical workflow [18]. In preceding publications two approaches were presented to automate this comparison [19][20][21]. These studies showed that the detection of ion beam range differences of 4 mm and more is possible in head and neck patients.…”

“…caused by a suboptimal determination of the range in human tissues [24]. So far, the investigations of the automated evaluation approaches were primarily relying on artificially altered patient data [19][20][21]. An objective reference data set to test the algorithms was so far not available although it is important to test the developed evaluation approaches on a set of reproducible PT-PET images that are not influenced by rather unpredictable patient depended factors.…”