ESTRO 2021 Abstract Book

S1553

ESTRO 2021

performed using NiftyReg, focused on the treated breast and ipsilateral lymph node regions, using either sum square difference (SSD) or normalised mutual information (NMI) cost function. Each patient was registered to Ref 0 , and to Ref i of the cluster it belongs to. The optimal number of reference patients was determined by assessing the performance of the image registration for each approach using normalised correlation coefficient (NCC) between the reference image and the deformed images. Differences were assessed using Wilcoxon test with a 95% confidence interval in SPSS version 18. Results Clusters and Silhouette analysis results are shown in Figure 1A. According to Silhouette analysis, the optimal number of clusters is 2. Patient characteristics of each cluster are plotted in Figure 1B, showing that patients are mostly clustered on BMI. Figure 2 shows the NCC for all registration methods. NMI-based DIR yielded superior results (higher NCC) than SSD- based DIR. Using 2 clusters (2 reference patients) did not improve the image registration result compared with complete cohort using affine ( p =0.397) or NMI ( p =0.533) but decreased the performance of SSD-based DIR ( p < 0.001) when considering the NCC value.

Conclusion Based on our results, a single reference patient is sufficient for accurate spatial normalization using NMI-based registration. The full cohort and patients with prone position will be included in future work.

Digital Poster: Treatment planning: optimisation and algorithms

PO-1823 Evaluation of dosimetric parameters with Co-60 and Ir-192 HDR brachytherapy of cervical cancer S. Srivastava 1 , N. Moirangthem 2 , N. Singh 1 1 King George's Medical University, Department of Radiotherapy, Lucknow, India; 2 Dr. B. Borooah Cancer Institute, Department of Radiation Oncology, Guwahati, India

Purpose or Objective To determine the differences in dosimetric parameters of treatment plans in patients treated with interstitial and