Knowledge-based dose prediction models to inform gynecologic brachytherapy needle supplementation for locally advanced cervical cancer

Improving plan quality in cervical brachytherapy using a simple knowledge-based prediction tool for OAR dose (D2cm3)

PURPOSE

Toxicity from cervical brachytherapy has been demonstrated to correlate with the D2cm3 of the bladder, rectum, and bowel. This suggests a simplified version of knowledge-based planning investigating the relationship of the overlap distance for 2cm3 and the D2cm3 from planning may be possible. This work demonstrates the feasibility of simple knowledge-based planning to predict the D2cm3, detect suboptimal plans, and improve plan quality.

METHODS AND MATERIALS

The overlap volume histogram (OVH) method was used to determine the distance for 2cm3 of overlap between the OAR and CTV_HR. Linear plots modeled the OAR D2cm3 and 2cm3 overlap distance. Two datasets of 20 patients (plans from 43 insertions in each dataset) were used to create two independent models, and the performance of each model was compared using cross-validation. Doses were scaled to ensure consistent CTV_HR D90 values. The predicted D2cm3 is entered as the maximum constraint in the inverse planning algorithm.

RESULTS

Mean bladder D2cm3 decreased by 2.9% for the models from each dataset, mean rectal D2cm3 decreased 14.9% for the model from dataset 1 and 6.0% for the model from dataset 2, mean sigmoid D2cm3 decreased 10.7% for the model from dataset 1 and 6.1% for the model from dataset 2, mean bowel D2cm3 decreased 4.1% for the model from dataset 1 but no statistically significant difference was observed for the model from dataset 2.

CONCLUSIONS

A simplified knowledge-based planning method was used to predict D2cm3 and was able to automate optimization of brachytherapy plans for locally advanced cervical cancer.

Automated treatment planning framework for brachytherapy of cervical cancer using 3D dose predictions

Objective. To lay the foundation for automated knowledge-based brachytherapy treatment planning using 3D dose estimations, we describe an optimization framework to convert brachytherapy dose distributions directly into dwell times (DTs).Approach. A dose rate kernelḋ(r,θ,φ)was produced by exporting 3D dose for one dwell position from the treatment planning system and normalizing by DT. By translating and rotating this kernel to each dwell position, scaling by DT and summing over all dwell positions, dose was computed (Dcalc). We used a Python-coded COBYLA optimizer to iteratively determine the DTs that minimize the mean squared error betweenDcalcand reference doseDref, computed using voxels withDref80%-120% of prescription. As validation of the optimization, we showed that the optimizer replicates clinical plans whenDref= clinical dose in 40 patients treated with tandem-and-ovoid (T&O) or tandem-and-ring (T&R) and 0-3 needles. Then we demonstrated automated planning in 10 T&O usingDref= dose predicted from a convolutional neural network developed in past work. Validation and automated plans were compared to clinical plans using mean absolute differences (MAD=1N∑n=1Nabsxn-xn') over all voxels (xn= Dose,N= #voxels) and DTs (xn= DT,N= #dwell positions), mean differences (MD) in organD2ccand high-risk CTV D90 over all patients (where positive indicates higher clinical dose), and mean Dice similarity coefficients (DSC) for 100% isodose contours.Main results. Validation plans agreed well with clinical plans (MADdose= 1.1%, MADDT= 4 s or 0.8% of total plan time,D2ccMD = -0.2% to 0.2% and D90 MD = -0.6%, DSC = 0.99). For automated plans, MADdose= 6.5% and MADDT= 10.3 s (2.1%). The slightly higher clinical metrics in automated plans (D2ccMD = -3.8% to 1.3% and D90 MD = -5.1%) were due to higher neural network dose predictions. The overall shape of the automated dose distributions were similar to clinical doses (DSC = 0.91).Significance. Automated planning with 3D dose predictions could provide significant time savings and standardize treatment planning across practitioners, regardless of experience.

Brachytherapy outcome modeling in cervical cancer patients: A predictive machine learning study on patient-specific clinical, physical and dosimetric parameters

PURPOSE

To predict clinical response in locally advanced cervical cancer (LACC) patients by a combination of measures, including clinical and brachytherapy parameters and several machine learning (ML) approaches.

METHODS

Brachytherapy features such as insertion approaches, source metrics, dosimetric, and clinical measures were used for modeling. Four different ML approaches, including LASSO, Ridge, support vector machine (SVM), and Random Forest (RF), were applied to extracted measures for model development alone or in combination. Model performance was evaluated using the area under the curve (AUC) of receiver operating characteristics curve, sensitivity, specificity, and accuracy. Our results were compared with a reference model developed by simple logistic regression applied to three distinct clinical features identified by previous papers.

RESULTS

One hundred eleven LACC patients were included. Nine data sets were obtained based on the features, and 36 predictive models were built. In terms of AUC, the model developed using RF applied to dosimetric, physical, and total BT sessions features were found as the most predictive [AUC; 0.82 (0.95 confidence interval (CI); 0.79 -0.93), sensitivity; 0.79, specificity; 0.76, and accuracy; 0.77]. The AUC (0.95 CI), sensitivity, specificity, and accuracy for the reference model were found as 0.56 (0.52 ...0.68), 0.51, 0.51, and 0.48, respectively. Most RF models had significantly better performance than the reference model (Bonferroni corrected p-value < 0.0014).

CONCLUSION

Brachytherapy response can be predicted using dosimetric and physical parameters extracted from treatment parameters. Machine learning algorithms, including Random Forest, could play a critical role in such predictive modeling.

Knowledge-based three-dimensional dose prediction for tandem-and-ovoid brachytherapy

PURPOSE

The purpose of this work was to develop a knowledge-based dose prediction system using a convolution neural network (CNN) for cervical brachytherapy treatments with a tandem-and-ovoid applicator.

METHODS

A 3D U-NET CNN was utilized to make voxel-wise dose predictions based on organ-at-risk (OAR), high-risk clinical target volume (HRCTV), and possible source location geometry. The model comprised 395 previously treated cases: training (273), validation (61), test (61). To assess voxel prediction accuracy, we evaluated dose differences in all cohorts across the dose range of 20-130% of prescription, mean(SD) and standard deviation (σ), as well as isodose dice similarity coefficients for clinical and/or predicted dose distributions. We examined discrete Dose-Volume Histogram(DVH) metrics utilized for brachytherapy plan quality assessment (HRCTV D90%, and bladder and/or rectum and/or sigmoid D2cc) with ΔD_x=D_x,actual-D_x,predicted Pearson correlation coefficient, standard deviation, and mean further quantifying model performance.

RESULTS

Ranges of voxel-wise dose difference accuracy (δD¯±σ) for 20-130% dose interval in training (test) sets ranged from [-0.5% ± 2.0% to +2.0% ± 14.0%] ([-0.1% ± 4.0% to +4.0% ± 26.0%]) in all voxels, [-1.7% ± 5.1% to -3.5% ± 12.8%] ([-2.9% ± 4.8% to -2.6% ± 18.9%]) in HRCTV, [-0.02% ± 2.40% to +3.2% ± 12.0%] ([-2.5% ± 3.6% to +0.8% ± 12.7%]) in bladder, [-0.7% ± 2.4% to +15.5% ± 11.0%] ([-0.9% ± 3.2% to +27.8% ± 11.6%]) in rectum, and [-0.7% ± 2.3% to +10.7% ± 15.0%] ([-0.4% ± 3.0% to +18.4% ± 11.4%]) in sigmoid. Isodose dice similarity coefficients ranged from [0.96,0.91] for training and [0.94,0.87] for test cohorts. Relative DVH metric prediction in the training (test) set were HRCTV ΔD¯₉₀±σ_ΔD=-0.19 ± 0.55Gy(-0.09 ± 0.67 Gy), bladder ΔD¯_2cc±σ_ΔD= -0.06 ± 0.54Gy(-0.17 ± 0.67 Gy), rectum ΔD¯_2cc±σ_ΔD= -0.03 ± 0.36Gy(-0.04 ± 0.46 Gy), and sigmoid ΔD¯_2cc±σ_ΔD= -0.01 ± 0.34Gy(0.00 ± 0.44 Gy).

CONCLUSIONS

A 3D knowledge-based dose predictions provide voxel-level and DVH metric estimates that could be used for treatment plan quality control and data-driven plan guidance.