ABSTRACT
CONTENTS 22.1 Introduction 368 22.2 Overall System Structure 369 22.3GPU Modules in the TARGET System 371
22.3.1Fluence Map Generation 371 22.3.2Monte Carlo Dose Engine 372 22.3.3GPU-Based MC Denoising Algorithm 373 22.3.4GPU-Based Gamma-Index Evaluation 374
22.4Experimental Results 375 22.4.1Dose Denoising 375 22.4.2Overall System Performance 376
22.5 Conclusion 377 References 380
Modern intensity-modulated radiation therapy (IMRT) and volumetric-modulated arc therapy (VMAT) technologies deliver more conformal dose to targets while sparing normal healthy tissues than conventional radiotherapy. To achieve these goals, the treatment planning and delivery process for IMRT and VMAT have become much more complicated and less intuitive to users. For instance, inverse treatment planning is utilized where values of thousands of variables, if not more, are adjusted by a computer program via specialized optimization algorithms to achieve the desired dosimetric goal. During delivery, many components in a linear accelerator (linac) are precisely controlled to deliver the optimized plan to the patient. is process is more error prone than conventional radiotherapy and consequences caused by the errors are probably more severe. erefore, a quality assurance (QA) procedure is needed before the rst treatment fraction to check for potential errors in the patient-specic plan, both in the treatment planning stage and in the plan delivery stage.
