ESTRO 2024 - Abstract Book

S4572

Physics - Machine learning models and clinical applications

ESTRO 2024

Male

Female

Male

Female

DL

0.94

0.95

0

0

0.22

Female reference

SyN

0.76

0.82

16

3

<0.01

DL

0.93

0.95

1

0

0.18

Male reference

SyN

0.72

0.77

14

9

0.18

Table 1. Quantitative results for DL-based and SyN registration stratified according to gender. Abbreviations: DL: deep learning, SyN: symmetric diffeomorphic image registration, DSC: dice similarity coefficient

Conclusion:

Our novel DL-based lung cancer registration method significantly reduced patient exclusions due to unsuccessful registration, compared to a conventional registration tool. This is clinically significant for VBA, as statistical power is directly related to patient size.

Keywords: registration, AI, voxel-based analysis

References:

[1] H. Wilson, G. Dervenoulas, and M. Politis, “Chapter Nine - Structural Magnetic Resonance Imaging in Huntington’s Disease,” in International Review of Neurobiology, vol. 142, M. Politis, Ed., in Imaging in Movement Disorders: Imaging in Atypical Parkinsonism and Familial Movement Disorders, vol. 142. , Academic Press, 2018, pp. 335–380. doi: 10.1016/bs.irn.2018.09.006. [2] J. Jiang and H. Veeraraghavan, “One shot PACS: Patient specific Anatomic Context and Shape prior aware recurrent registration-segmentation of longitudinal thoracic cone beam CTs,” IEEE Trans. Med. Imaging, vol. 41, no. 8, pp. 2021–2032, Aug. 2022, doi: 10.1109/TMI.2022.3154934.

3161

Mini-Oral

A novel knowledge-based planning pipeline for generating Gamma Knife treatment plans

Binghao Zhang 1 , Aaron Babier 1 , Mark Ruschin 2 , Timothy C.Y. Chan 1

1 University of Toronto, Mechanical and Industrial Engineering, Toronto, Canada. 2 Sunnybrook Health Sciences Center, Odette Cancer Centre, Toronto, Canada

Purpose/Objective: