Learning-Based Stopping Power Mapping on Dual-Energy CT for Proton Radiation Therapy
Purpose: Dual-energy computed tomography (DECT) has been used to derive relative stopping power (RSP) maps by obtaining the energy dependence of photon interactions. The DECT-derived RSP maps could potentially be compromised by image noise levels and the severity of artifacts when using physics-base...
| Main Authors: | , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Particle Therapy Co-operative Group
2021-02-01
|
| Series: | International Journal of Particle Therapy |
| Subjects: | |
| Online Access: | https://theijpt.org/doi/pdf/10.14338/IJPT-D-20-00020.1 |
| _version_ | 1797203267099295744 |
|---|---|
| author | Tonghe Wang, PhD Yang Lei, PhD Joseph Harms, PhD Beth Ghavidel, MS Liyong Lin, PhD Jonathan J. Beitler, MD Mark McDonald, MD Walter J. Curran, MD Tian Liu, PhD Jun Zhou, PhD Xiaofeng Yang, PhD |
| author_facet | Tonghe Wang, PhD Yang Lei, PhD Joseph Harms, PhD Beth Ghavidel, MS Liyong Lin, PhD Jonathan J. Beitler, MD Mark McDonald, MD Walter J. Curran, MD Tian Liu, PhD Jun Zhou, PhD Xiaofeng Yang, PhD |
| author_sort | Tonghe Wang, PhD |
| collection | DOAJ |
| description | Purpose: Dual-energy computed tomography (DECT) has been used to derive relative stopping power (RSP) maps by obtaining the energy dependence of photon interactions. The DECT-derived RSP maps could potentially be compromised by image noise levels and the severity of artifacts when using physics-based mapping techniques. This work presents a noise-robust learning-based method to predict RSP maps from DECT for proton radiation therapy.
Materials and Methods: The proposed method uses a residual attention cycle-consistent generative adversarial network to bring DECT-to-RSP mapping close to a 1-to-1 mapping by introducing an inverse RSP-to-DECT mapping. To evaluate the proposed method, we retrospectively investigated 20 head-and-neck cancer patients with DECT scans in proton radiation therapy simulation. Ground truth RSP values were assigned by calculation based on chemical compositions and acted as learning targets in the training process for DECT datasets; they were evaluated against results from the proposed method using a leave-one-out cross-validation strategy.
Results: The predicted RSP maps showed an average normalized mean square error of 2.83% across the whole body volume and an average mean error less than 3% in all volumes of interest. With additional simulated noise added in DECT datasets, the proposed method still maintained a comparable performance, while the physics-based stoichiometric method suffered degraded inaccuracy from increased noise level. The average differences from ground truth in dose volume histogram metrics for clinical target volumes were less than 0.2 Gy for D95% and Dmax with no statistical significance. Maximum difference in dose volume histogram metrics of organs at risk was around 1 Gy on average.
Conclusion: These results strongly indicate the high accuracy of RSP maps predicted by our machine-learning–based method and show its potential feasibility for proton treatment planning and dose calculation. |
| first_indexed | 2024-04-24T08:16:37Z |
| format | Article |
| id | doaj.art-ef918afd2a69455eaad220367c6604f2 |
| institution | Directory Open Access Journal |
| issn | 2331-5180 |
| language | English |
| last_indexed | 2024-04-24T08:16:37Z |
| publishDate | 2021-02-01 |
| publisher | Particle Therapy Co-operative Group |
| record_format | Article |
| series | International Journal of Particle Therapy |
| spelling | doaj.art-ef918afd2a69455eaad220367c6604f22024-04-17T03:28:17ZengParticle Therapy Co-operative GroupInternational Journal of Particle Therapy2331-51802021-02-0173466010.14338/IJPT-D-20-00020.12331-5180-7-3-46Learning-Based Stopping Power Mapping on Dual-Energy CT for Proton Radiation TherapyTonghe Wang, PhD0Yang Lei, PhD1Joseph Harms, PhD2Beth Ghavidel, MS3Liyong Lin, PhD4Jonathan J. Beitler, MD5Mark McDonald, MD6Walter J. Curran, MD7Tian Liu, PhD8Jun Zhou, PhD9Xiaofeng Yang, PhD10Department of Radiation Oncology and Winship Cancer Institute, Emory University, Atlanta, GA, USADepartment of Radiation Oncology and Winship Cancer Institute, Emory University, Atlanta, GA, USADepartment of Radiation Oncology and Winship Cancer Institute, Emory University, Atlanta, GA, USADepartment of Radiation Oncology and Winship Cancer Institute, Emory University, Atlanta, GA, USADepartment of Radiation Oncology and Winship Cancer Institute, Emory University, Atlanta, GA, USADepartment of Radiation Oncology and Winship Cancer Institute, Emory University, Atlanta, GA, USADepartment of Radiation Oncology and Winship Cancer Institute, Emory University, Atlanta, GA, USADepartment of Radiation Oncology and Winship Cancer Institute, Emory University, Atlanta, GA, USADepartment of Radiation Oncology and Winship Cancer Institute, Emory University, Atlanta, GA, USADepartment of Radiation Oncology and Winship Cancer Institute, Emory University, Atlanta, GA, USADepartment of Radiation Oncology and Winship Cancer Institute, Emory University, Atlanta, GA, USAPurpose: Dual-energy computed tomography (DECT) has been used to derive relative stopping power (RSP) maps by obtaining the energy dependence of photon interactions. The DECT-derived RSP maps could potentially be compromised by image noise levels and the severity of artifacts when using physics-based mapping techniques. This work presents a noise-robust learning-based method to predict RSP maps from DECT for proton radiation therapy. Materials and Methods: The proposed method uses a residual attention cycle-consistent generative adversarial network to bring DECT-to-RSP mapping close to a 1-to-1 mapping by introducing an inverse RSP-to-DECT mapping. To evaluate the proposed method, we retrospectively investigated 20 head-and-neck cancer patients with DECT scans in proton radiation therapy simulation. Ground truth RSP values were assigned by calculation based on chemical compositions and acted as learning targets in the training process for DECT datasets; they were evaluated against results from the proposed method using a leave-one-out cross-validation strategy. Results: The predicted RSP maps showed an average normalized mean square error of 2.83% across the whole body volume and an average mean error less than 3% in all volumes of interest. With additional simulated noise added in DECT datasets, the proposed method still maintained a comparable performance, while the physics-based stoichiometric method suffered degraded inaccuracy from increased noise level. The average differences from ground truth in dose volume histogram metrics for clinical target volumes were less than 0.2 Gy for D95% and Dmax with no statistical significance. Maximum difference in dose volume histogram metrics of organs at risk was around 1 Gy on average. Conclusion: These results strongly indicate the high accuracy of RSP maps predicted by our machine-learning–based method and show its potential feasibility for proton treatment planning and dose calculation.https://theijpt.org/doi/pdf/10.14338/IJPT-D-20-00020.1stopping powerdual-energy ctproton therapymachine learning |
| spellingShingle | Tonghe Wang, PhD Yang Lei, PhD Joseph Harms, PhD Beth Ghavidel, MS Liyong Lin, PhD Jonathan J. Beitler, MD Mark McDonald, MD Walter J. Curran, MD Tian Liu, PhD Jun Zhou, PhD Xiaofeng Yang, PhD Learning-Based Stopping Power Mapping on Dual-Energy CT for Proton Radiation Therapy International Journal of Particle Therapy stopping power dual-energy ct proton therapy machine learning |
| title | Learning-Based Stopping Power Mapping on Dual-Energy CT for Proton Radiation Therapy |
| title_full | Learning-Based Stopping Power Mapping on Dual-Energy CT for Proton Radiation Therapy |
| title_fullStr | Learning-Based Stopping Power Mapping on Dual-Energy CT for Proton Radiation Therapy |
| title_full_unstemmed | Learning-Based Stopping Power Mapping on Dual-Energy CT for Proton Radiation Therapy |
| title_short | Learning-Based Stopping Power Mapping on Dual-Energy CT for Proton Radiation Therapy |
| title_sort | learning based stopping power mapping on dual energy ct for proton radiation therapy |
| topic | stopping power dual-energy ct proton therapy machine learning |
| url | https://theijpt.org/doi/pdf/10.14338/IJPT-D-20-00020.1 |
| work_keys_str_mv | AT tonghewangphd learningbasedstoppingpowermappingondualenergyctforprotonradiationtherapy AT yangleiphd learningbasedstoppingpowermappingondualenergyctforprotonradiationtherapy AT josephharmsphd learningbasedstoppingpowermappingondualenergyctforprotonradiationtherapy AT bethghavidelms learningbasedstoppingpowermappingondualenergyctforprotonradiationtherapy AT liyonglinphd learningbasedstoppingpowermappingondualenergyctforprotonradiationtherapy AT jonathanjbeitlermd learningbasedstoppingpowermappingondualenergyctforprotonradiationtherapy AT markmcdonaldmd learningbasedstoppingpowermappingondualenergyctforprotonradiationtherapy AT walterjcurranmd learningbasedstoppingpowermappingondualenergyctforprotonradiationtherapy AT tianliuphd learningbasedstoppingpowermappingondualenergyctforprotonradiationtherapy AT junzhouphd learningbasedstoppingpowermappingondualenergyctforprotonradiationtherapy AT xiaofengyangphd learningbasedstoppingpowermappingondualenergyctforprotonradiationtherapy |