Deep learning based uncertainty prediction of deformable image registration for contour propagation and dose accumulation in online adaptive radiotherapy

Abstract

Objective: Online adaptive radiotherapy aims to fully leverage the advantages of highly conformal therapy by reducing anatomical and set-up uncertainty, thereby alleviating the need for robust treatments. This requires extensive automation, among which are the use of deformable image registration (DIR) for contour propagation and dose accumulation. However, inconsistencies in DIR solutions between different algorithms have caused distrust, hampering its direct clinical use. This work aims to enable the clinical use of DIR by developing deep learning methods to predict DIR uncertainty and propagating it into clinically usable metrics.

Approach: Supervised and unsupervised neural networks were trained to predict the Gaussian uncertainty of a given deformable vector field (DVF). Since both methods rely on different assumptions, their predictions differ and were further merged into a combined model. The resulting normally distributed DVFs can be directly sampled to propagate the uncertainty into contour and accumulated dose uncertainty.

Main results: The unsupervised and combined models can accurately predict the uncertainty in the manually annotated landmarks on the DIRLAB dataset. Furthermore, for 5 patients with lung cancer, the propagation of the predicted DVF uncertainty into contour uncertainty yielded for both methods an expected calibration error of less than 3%. Additionally, the probabilisticly accumulated dose volume histograms (DVH) encompass well the accumulated proton therapy doses using 5 different DIR algorithms. It was additionally shown that the unsupervised model can be used for different DIR algorithms without the need for retraining. 

Significance: Our work presents first-of-a-kind deep learning methods to predict the uncertainty of the DIR process. The methods are fast, yield high-quality uncertainty estimates and are useable for different algorithms and applications. This allows clinics to use DIR uncertainty in their workflows without the need to change their DIR implementation.

Artificial Intelligence Uncertainty Quantification in Radiotherapy Applications - A Scoping Review

“…Subsequently, carefully designed UQ may help identify patients for whom the model’s predictions are more reliable. Finally, despite the importance of using diverse and heterogeneous data for uncertainty experiments, particularly for determining how well models handle new and unknown data scenarios [22], only a handful of studies attempted to utilize multiple external test datasets [48,57,88,92]. Interestingly, this was in stark contrast to a previous scoping review on AI UQ in a broader medical context which identified a predominance of external dataset testing [31].…”

A review of the clinical introduction of 4D particle therapy research concepts

“…However, also the field of PT shows broad interest in implementing more online or hybrid approaches, especially including daily replanning and employing plan libraries [8] , [22] , [126] , [127] , [128] . Recently, several key players in the field of real-time adaptive PT joined forces [129] to tackle the remaining challenges and pave the way towards online adaptation in PT [130] , [131] , [132] , [133] , [134] , [135] , [136] . These challenges include aspects of dose accumulation, contour propagation, daily plan re-optimisation and approval as well as quality assurance of the online adaptive treatments [137] , [138] .…”

Review and recommendations on deformable image registration uncertainties for radiotherapy applications

“…The simplest method is using the worst-case or average difference distance in all directions for all voxels of a given region or structure. More individualised methods have been investigated (Amstutz et al 2021b, Smolders et al 2022b, 2023c and provide patient specific voxel-wise uncertainties maps. We recommend using such voxel-wise uncertainty maps whenever possible.…”