Hardware acceleration of automated 4DCT analysis

Robinson, Fraser and Crockett, Louise and Nailon, Bill and Stewart, Bob and McLaren, Duncan (2017) Hardware acceleration of automated 4DCT analysis. In: 6th Annual Scientific Meeting of the Scottish Radiotherapy Research Forum, 2017-11-02 - 2017-11-02, Stirling Court Hotel (University of Stirling).

[thumbnail of Robinson-etal-ScoRRF-2017-Hardware-acceleration-of-automated-4DCT-analysis]
Preview
Text. Filename: Robinson_etal_ScoRRF_2017_Hardware_acceleration_of_automated_4DCT_analysis.pdf
Accepted Author Manuscript
License: Creative Commons Attribution 4.0 logo

Download (1MB)| Preview

Abstract

Background -- Stereotactic ablative radiotherapy (SABR) requires cancerous lesions to be more accurately targeted than conventional techniques. The time interval in which a treatment fraction is delivered precludes manual delineation of the gross tumour volume (GTV) and organs at risk (OAR) to correct for intrafraction motion. However, automatic segmentation techniques may be able to achieve this. The runtime performance of image analysis algorithms can benefit from implementation on appropriate hardware architectures. Field Programmable Gate Arrays (FPGA), which contain customisable hardware, have the potential to enable real-time image processing. Aims/Objectives -- The aim of this study was to develop an FPGA-based approach for automatic image segmentation of 4DCT scans. The performance of the algorithm was assessed in terms of the accuracy of the segmentation and the runtime performance. Methods/Results -- The segmentation algorithm was based on Otsu’s method and measured the range of motion of a phantom in eight 4DCT scans. The algorithm was implemented on an FPGA-based platform and a CPU to compare the runtime performance. The detected range of motion was accurate in seven cases and in the eighth case, was inaccurate by the CT slice thickness. The FPGA-based implementation executed in 14.8ms, around 14% faster than on the CPU. Conclusions -- This study demonstrates the ability of hardware-accelerated image processing algorithms to aid radiotherapy. This work detected ranges of motion of a phantom, but could be extended to consider clinical imaging data. It is intended to extend this work by using the FPGA device to accelerate algorithms to perform real-time adaptive radiotherapy.