Methodological approach to create an atlas using a commercial auto-contouring software.

Marta Casati,Lorenzo Livi,Silvia Calusi,Isacco Desideri,Vanessa Di Cataldo,Stefania Pallotta,Marco Vernaleone,Gabriele Simontacchi,Livia Marrazzo,Stefano Piffer,Giulio Francolini,Daniela Greto

doi:10.1002/acm2.13093

Abstract

PurposeThe aim of this work was to establish a methodological approach for creation and optimization of an atlas for auto‐contouring, using the commercial software MIM MAESTRO (MIM Software Inc. Cleveland OH).MethodsA computed tomography (CT) male pelvis atlas was created and optimized to evaluate how different tools and options impact on the accuracy of automatic segmentation. Pelvic lymph nodes (PLN), rectum, bladder, and femurs of 55 subjects were reviewed for consistency by a senior consultant radiation oncologist with 15 yr of experience. Several atlas and workflow options were tuned to optimize the accuracy of auto‐contours. The deformable image registration (DIR), the finalization method, the k number of atlas best matching subjects, and several post‐processing options were studied. To test our atlas performances, automatic and reference manual contours of 20 test subjects were statistically compared based on dice similarity coefficient (DSC) and mean distance to agreement (MDA) indices. The effect of field of view (FOV) reduction on auto‐contouring time was also investigated.ResultsWith the optimized atlas and workflow, DSC and MDA median values of bladder, rectum, PLN, and femurs were 0.91 and 1.6 mm, 0.85 and 1.6 mm, 0.85 and 1.8 mm, and 0.96 and 0.5 mm, respectively. Auto‐contouring time was more than halved by strictly cropping the FOV of the subject to be contoured to the pelvic region.ConclusionA statistically significant improvement of auto‐contours accuracy was obtained using our atlas and optimized workflow instead of the MIM Software pelvic atlas.

Highlights

In radiotherapy planning, image segmentation is one of the preliminary and time‐consuming tasks, affected by interobserver variability.[1,2,3,4,5,6] This procedure, usually performed on computed tomography (CT) images, is affected by the scarce image contrast that hinders the application of semi‐automatic segmentation algorithms based on threshold or region growing
The comparison between reference and auto‐contours obtained using Atlas 1 and Atlas 2 is reported in Fig. 2, showing dice similarity coefficient (DSC) and mean distance to agreement (MDA) box plot evaluated for each patient and region of interest (ROI)
DSC and MDA similarity indices are reported in box plots of Fig. 3 for each ROI and for three smoothing factors for the Same‐subject algorithm

Summary

Introduction

Image segmentation is one of the preliminary and time‐consuming tasks, affected by interobserver variability.[1,2,3,4,5,6] This procedure, usually performed on computed tomography (CT) images, is affected by the scarce image contrast that hinders the application of semi‐automatic segmentation algorithms based on threshold or region growing. Some of the criteria to define targets and organs at risk (OAR) are not related to CT visible anatomical boundaries. For these reasons, image segmentation for radiotherapy treatment planning is still a challenging and labor‐ intensive task. The atlas‐based approach relies on the availability of one or more CT series of a certain anatomical district, already contoured by an expert physician following guidelines.[7,8] The operating principle is to perform a deformable registration of an atlas subject on the new subject and apply the same transformation to the atlas structures, to obtain a proposal of contouring for the new subject. Each subject of the multisubject atlas could be deformed on the subject to be contoured, to obtain N possible sets of structures. From each of the k subjects, a contours proposal is derived, and a finalization algorithm combines these k series of contours into a single set of contours

Objectives

Methods

Results

Discussion

Conclusion