Deep learning-based dose prediction to enhance planning efficiency in cervical brachytherapy with hybrid applicators

Clinical Snapshot

Key Findings

• P Value: p > 0.05 for all key DVH parameter comparisons between predicted and manually generated plans (non-inferiority direction)

• Effect Size: MAE of 0.45 ± 0.27 Gy for voxel-wise dose prediction; small dosimetric differences observed for HR-CTV D98%, HR-CTV D90%, and bladder D2cc between predicted and clinical plans

• Primary Outcome: Voxel-wise dose prediction accuracy assessed by mean absolute error (MAE) on the 40-plan testing dataset, and DVH parameter agreement on the 15-plan clinical evaluation dataset

• Nnt Or Sensitivity: DSC values of 0.74–0.77 for isodose volume spatial agreement (moderate spatial overlap; DSC ≥ 0.80 is generally considered good agreement in radiotherapy planning literature)

• Confidence Interval: Not reported

Clinical Application

Feasibility as a clinical decision-support tool is plausible but not yet demonstrated prospectively. Integration into treatment planning system workflows would require software validation, regulatory clearance, and staff training. The model would function as a reference dose distribution to guide dwell time optimisation rather than as a fully automated planning solution. Cervical cancer brachytherapy is performed at major Australian radiation oncology centres including Peter MacCallum Cancer Centre, Royal Brisbane and Women's Hospital, and Chris O'Brien Lifehouse. Hybrid applicator brachytherapy is increasingly adopted in line with GEC-ESTRO recommendations. The TGA would require regulatory approval for any AI-based dose planning tool used clinically in Australia. RANZCR and ACPSEM would need to endorse clinical implementation guidelines. PBS funding does not directly apply to brachytherapy planning software, but MBS item numbers cover brachytherapy procedures. Australian centres would require local retraining or fine-tuning of the model on institutional data before deployment, given the single-institution Thai training dataset. Patients with cervical cancer undergoing HDR intracavitary-interstitial brachytherapy using hybrid applicators at centres with established adaptive brachytherapy workflows and MRI-guided contouring capability

Abstract

Objective.This study aimed to develop a dose prediction model for hybrid applicators in cervical cancer brachytherapy.Approach.A total of 216 previous treatment plans were collected and divided into three datasets: training (161 plans), testing (40 plans), and clinical evaluation (15 plans). The treatment plans included three hybrid applicator types: tandem and ovoids combined with interstitial needles, tandem and ring combined with interstitial needles, and tandem and cylinder combined with interstitial needles. The model was validated on the testing dataset using mean absolute error (MAE) for voxel-wise dose evaluation and on the clinical evaluation dataset using dice similarity coefficient (DSC) and dose statistic differences.Main results.The model achieved a voxel-wise MAE of 0.45 ± 0.27 Gy and showed no statistically significant differences (p> 0.05) in all key dose-volume histogram parameters when compared to clinical plans. The DSC values ranged from 0.74 to 0.77. Small dosimetric differences between the optimized and manually generated plans were observed for the high-risk clinical target volume (HR-CTV) D98%, HR-CTV D90%, and bladder D2cc.Significance.The developed dose prediction model demonstrates significant potential for application in hybrid applicator brachytherapy. This model can provide guidance for dose delivery during dwell time optimization. As a reference for optimization, it has the potential to streamline the planning process, reduce planning time, and improve consistency in adaptive workflows.

References

1. 1.Thongsawad, S., Rinjan, N., Tharavichitkul, E., Khamfongkhruea, C., Tannanonta, C., & Chanpanya, T. (2026). Deep learning-based dose prediction to enhance planning efficiency in cervical brachytherapy with hybrid applicators. Physics in Medicine and Biology. https://doi.org/10.1088/1361-6560/ae7cd4