Publications by authors named "Benjamin P Ziemer"

4 Publications

  • Page 1 of 1

Attention-Aware Discrimination for MR-to-CT Image Translation Using Cycle-Consistent Generative Adversarial Networks.

Radiol Artif Intell 2020 Mar 25;2(2):e190027. Epub 2020 Mar 25.

Department of Radiation Oncology, University of California, 1600 Divisidero St, San Francisco, CA 94115.

Purpose: To suggest an attention-aware, cycle-consistent generative adversarial network (A-CycleGAN) enhanced with variational autoencoding (VAE) as a superior alternative to current state-of-the-art MR-to-CT image translation methods.

Materials And Methods: An attention-gating mechanism is incorporated into a discriminator network to encourage a more parsimonious use of network parameters, whereas VAE enhancement enables deeper discrimination architectures without inhibiting model convergence. Findings from 60 patients with head, neck, and brain cancer were used to train and validate A-CycleGAN, and findings from 30 patients were used for the holdout test set and were used to report final evaluation metric results using mean absolute error (MAE) and peak signal-to-noise ratio (PSNR).

Results: A-CycleGAN achieved superior results compared with U-Net, a generative adversarial network (GAN), and a cycle-consistent GAN. The A-CycleGAN averages, 95% confidence intervals (CIs), and Wilcoxon signed-rank two-sided test statistics are shown for MAE (19.61 [95% CI: 18.83, 20.39], = .0104), structure similarity index metric (0.778 [95% CI: 0.758, 0.798], = .0495), and PSNR (62.35 [95% CI: 61.80, 62.90], = .0571).

Conclusion: A-CycleGANs were a superior alternative to state-of-the-art MR-to-CT image translation methods.© RSNA, 2020.
View Article and Find Full Text PDF

Download full-text PDF

Source Listing
March 2020

Heuristic knowledge-based planning for single-isocenter stereotactic radiosurgery to multiple brain metastases.

Med Phys 2017 Oct 30;44(10):5001-5009. Epub 2017 Aug 30.

Department of Radiation Medicine and Applied Sciences, University of California, San Diego, La Jolla, CA, USA.

Purpose: Single-isocenter, volumetric-modulated arc therapy (VMAT) stereotactic radiosurgery (SRS) for multiple brain metastases (multimets) can deliver highly conformal dose distributions and reduce overall patient treatment time compared to other techniques. However, treatment planning for multimet cases is highly complex due to variability in numbers and sizes of brain metastases, as well as their relative proximity to organs-at-risk (OARs). The purpose of this study was to automate the VMAT planning of multimet cases through a knowledge-based planning (KBP) approach that adapts single-target SRS dose predictions to multiple target predictions.

Methods: Using a previously published artificial neural network (ANN) KBP system trained on single-target, linac-based SRS plans, 3D dose distribution predictions for multimet patients were obtained by treating each brain lesion as a solitary target and subsequently combining individual dose predictions into a single distribution. Spatial dose distributions di(r→) for each of the i = 1…N lesions were merged using the combination function d(r→)=∑iNdin(r→)1/n. The optimal value of n was determined by minimizing root-mean squared (RMS) difference between clinical multimet plans and predicted dose per unit length along the line profile joining each lesion in the clinical cohort. The gradient measure GM=[3/4π]1/3V50%1/3-V100%1/3 is the primary quality metric for SRS plan evaluation at our institution and served as the main comparative metric between clinical plans and the KBP results. A total of 41 previously treated multimet plans, with target numbers ranging from N = 2-10, were used to validate the ANN predictions and subsequent KBP auto-planning routine. Fully deliverable KBP plans were developed by converting predicted dose distribution into patient-specific optimization objectives for the clinical treatment planning system (TPS). Plan parity was maintained through identical arc configuration and target normalization. Overall plan quality improvements were quantified by calculating the difference between SRS quality metrics (QMs): ΔQM = QM  - QM . In addition to GM, investigated QMs were: volume of brain receiving ≥ 10 Gy (V ), volume of brain receiving ≥ 5 Gy (ΔV ), heterogeneity index (HI), dose to 0.1 cc of the brainstem (D ), dose to 1% of the optic chiasm (D ), and interlesion dose (D ). In addition to this quantitative analysis, overall plan quality was assessed via blinded plan comparison of the manual and KBP treatment plans by SRS-specializing physicians.

Results: A dose combination factor of n = 8 yielded an integrated dose profile RMS difference of 2.9% across the 41-patient cohort. Multimet dose predictions exhibited ΔGM = 0.07 ± 0.10 cm against the clinical sample, implying either further normal tissue sparing was possible or that dose predictions were slightly overestimating achievable dose gradients. The latter is the more likely explanation, as this bias vanished when dose predictions were converted to deliverable KBP plans ΔGM = 0.00 ± 0.08 cm. Remaining QMs were nearly identical or showed modest improvements in the KBP sample. Equivalent QMs included: ΔV  = 0.37 ± 3.78 cc, ΔHI = 0.02 ± 0.08 and ΔD  = -2.22 ± 171.4 cGy. The KBP plans showed a greater degree of normal tissue sparing as indicated by brain ΔV  = 4.11± 24.05 cc, brainstem ΔD  = 42.8 ± 121.4 cGy, and chiasm ΔD  = 50.8 ± 83.0 cGy. In blinded review by SRS-specializing physicians, KBP-generated plans were deemed equivalent or superior in 32/41(78.1%) of the cases.

Conclusion: Heuristic KBP-driven automated planning in linac-based, single-isocenter treatments for multiple brain metastases maintained or exceeded overall plan quality.
View Article and Find Full Text PDF

Download full-text PDF

Source Listing
October 2017

Fully automated, comprehensive knowledge-based planning for stereotactic radiosurgery: Preclinical validation through blinded physician review.

Pract Radiat Oncol 2017 Nov - Dec;7(6):e569-e578. Epub 2017 Apr 19.

Department of Radiation Medicine and Applied Sciences, University of California, San Diego, La Jolla, California. Electronic address:

Purpose: As knowledge-based planning (KBP) attempts to augment and potentially supplant manual treatment planning, it is imperative to ensure any implementation maintains or improves overall plan quality in any disease site. The purpose of this study was to demonstrate the overall quality of KBP-driven automated stereotactic radiosurgery (SRS) treatment planning using blinded physician comparison and determine systematic factors predictive of physician plan preference to guide future KBP refinement.

Methods And Materials: Automated noncoplanar volume modulated arc therapy KBP routines were developed for 199 plans across 3 clinical SRS scenarios: isolated lesions (isolated), lesions closely abutting (<3 cm) organs at risk (involved), and single-isocenter multiple metastases (multimet). Overall plan quality and preference were assessed via blinded review of the plans by two SRS physicians. Quantitative quality metrics were also compared to determine systematic differences in the treatment plans. Multiple parameters were investigated as predictors of KBP plan selection.

Results: For the isolated, involved, and multimet scenarios, the KBP plans were considered to be superior or equivalent to clinical plans 86.7% (91/105), 81.1% (43/53), and 78.1% (32/41) of the time, respectively. All investigated quality metrics were equivalent or indicated more sparing for all KBP plans. The only nondosimetric predictor was planning target volume in the isolated (P = .02) and involved (P = .05) groups. The dosimetric predictors for the isolated group were gradient measure and heterogeneity index (both P < .01). In the multimet category, the only significant dosimetric predictor was interlesion dose (P = .01).

Conclusions: The fully automated KBP SRS plans were equivalent or superior to previously treated plans in 83.4% (166/199) of cases. In clinical implementation, geometric features found to be predictive of KBP performance can be used to identify plans where KBP results might benefit from further refinement, whereas dosimetric predictive features could be used to further refine KBP optimization priorities.
View Article and Find Full Text PDF

Download full-text PDF

Source Listing
July 2018

Dynamic CT perfusion measurement in a cardiac phantom.

Int J Cardiovasc Imaging 2015 Oct 9;31(7):1451-9. Epub 2015 Jul 9.

Department of Radiological Sciences, Medical Sciences I, B-140, University of California, Irvine, CA, 92697, USA.

Widespread clinical implementation of dynamic CT myocardial perfusion has been hampered by its limited accuracy and high radiation dose. The purpose of this study was to evaluate the accuracy and radiation dose reduction of a dynamic CT myocardial perfusion technique based on first pass analysis (FPA). To test the FPA technique, a pulsatile pump was used to generate known perfusion rates in a range of 0.96-2.49 mL/min/g. All the known perfusion rates were determined using an ultrasonic flow probe and the known mass of the perfusion volume. FPA and maximum slope model (MSM) perfusion rates were measured using volume scans acquired from a 320-slice CT scanner, and then compared to the known perfusion rates. The measured perfusion using FPA (P(FPA)), with two volume scans, and the maximum slope model (P(MSM)) were related to known perfusion (P(K)) by P(FPA) = 0.91P(K) + 0.06 (r = 0.98) and P(MSM) = 0.25P(K) - 0.02 (r = 0.96), respectively. The standard error of estimate for the FPA technique, using two volume scans, and the MSM was 0.14 and 0.30 mL/min/g, respectively. The estimated radiation dose required for the FPA technique with two volume scans and the MSM was 2.6 and 11.7-17.5 mSv, respectively. Therefore, the FPA technique can yield accurate perfusion measurements using as few as two volume scans, corresponding to approximately a factor of four reductions in radiation dose as compared with the currently available MSM. In conclusion, the results of the study indicate that the FPA technique can make accurate dynamic CT perfusion measurements over a range of clinically relevant perfusion rates, while substantially reducing radiation dose, as compared to currently available dynamic CT perfusion techniques.
View Article and Find Full Text PDF

Download full-text PDF

Source Listing
October 2015