Accurate functional volume definition in PET for radiotherapy treatment planning

Accurate volume contouring in PET is now considered crucial in radiotherapy as the use of functional imaging allows improved gross tumour volume (GTV) definition. On the other hand, an accurate delineation of the GTV as well as the definition of variable activity accumulation regions inside the tumour itself may facilitate the applications of “dose painting” for optimization of dosimetry. The objectives of such optimization include lower doses delivered to healthy surrounding tissues and higher doses delivered to malignant ones. Current state of the art algorithms for functional volume segmentation consist of adaptive threshold approaches. We have developed a segmentation approach for inhomogeneous tumours in PET, namely the FLAB (Fuzzy Locally Adaptive Bayesian), that was previously validated on simulated images. In this study, we investigated the accuracy of this algorithm in comparison to threshold-based approaches, applied to images of lung cancer patients scanned with FDG PET/CT. Simulated tumours were generated based on the activity distribution and shapes of the real lesions imaged on these patients in order to establish a “ground truth” as far as functional tumour volume is concerned. In addition, some of the patients were subsequently operated with the tumours removed and a subsequent macroscopic investigation performed to determine the true tumour sizes. These were compared to those from the segmented volumes obtained using the different algorithms under investigation. The FLAB algorithm is able to accurately extract the overall tumour from the healthy background tissues, as well as precisely delineate variable activity concentration regions of interest inside the tumours, whereas the other methodologies fail to do so. In addition, the FLAB results were systematically closer to the histology results than the other methodologies considered. Future studies will investigate the impact of the use of FLAB in radiotherapy treatmen- - t planning.