Automatic generation of three-dimensional dose reconstruction data for two-dimensional radiotherapy plans for historically treated patients

Wang, Ziyuan; Virgolin, Marco; Bosman, Peter A. N.; Crama, K.; Balgobind, Brian V.; Bel, Arjan; Alderliesten, Tanja

doi:10.1117/1.jmi.7.1.015001

Cited by 2 publications

(13 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our DRRs were generated in a conformal fashion, e.g., the abdomen was always fully included in RL direction. The automatic landmark detection that was used to generate features expects this conformity to achieve precise detection (Wang et al 2020). When dealing with actual historical radiographs, however, several challenges need to be taken into account.…”

Section: Discussionmentioning

confidence: 99%

“…The random sampling of the plan side led to 142 left-sided plans and 158 right-sided plans (roughly half-half). The same set of plan features used in our previous work was considered to generate plans in DICOM RTPLAN format (e.g., gantry and collimator angles, isocenter location, field sizes) (Wang et al 2020).…”

Section: Automatic Generation Of Artificial Wilms' Tumor Plansmentioning

confidence: 99%

“…Secondly, for each of the 42,600 plan emulations, dose-volume metrics of interest (see Sec. 2.3.1) were collected for the different OARs by use of our automatic dose computation pipeline (Wang et al 2020). The pipeline used the collapsed cone dose calculation algorithm of Oncentra TPS (version 4.3, Elekta AB, Stockhom, Sweden).…”

Section: Generation Of the Dataset For MLmentioning

confidence: 99%

“…For the features related to anatomical geometry, anatomical landmarks from DRRs were detected automatically using the landmark detection method in our previous work (Wang et al 2020). Note that the landmarks concerned only bony anatomy because other internal anatomy tissues are not reliably visible in historical radiographs.…”

Section: Explanatory Variables: Features Of Patients and Plansmentioning

confidence: 99%

“…Importantly, we normalized features related to measurements of anatomy and anatomy-plan geometry configuration (e.g., rib-cage width, field sizes, distances between landmarks and the isocenter) by the width and height of the respective DRR they were measured from (after the DRRs were cropped to a same region of interest between T10 and S1). This was done because when plans are emulated, they are scaled based on proportions derived from the landmarks (Wang et al 2018, Wang et al 2020. Since differences in anatomy solely due to overall anatomy scaling do not result in different dose-volume metric values, these differences should not be accounted for by the explanatory variables (confirmed in preliminary experiments).…”

Section: Explanatory Variables: Features Of Patients and Plansmentioning

confidence: 99%

See 4 more Smart Citations

Surrogate-free machine learning-based organ dose reconstruction for pediatric abdominal radiotherapy

Virgolin,

Wang,

Balgobind

et al. 2020

Preprint

View full text Add to dashboard Cite

To study radiotherapy-related adverse effects, detailed dose information (3D distribution) is needed for accurate dose-effect modeling. For childhood cancer survivors who underwent radiotherapy in the pre-CT era, only 2D radiographs were acquired, thus 3D dose distributions must be reconstructed from limited information. State-of-the-art methods achieve this by using 3D surrogate anatomies. These can however lack personalization and lead to coarse reconstructions. We present and validate a surrogate-free dose reconstruction method based on Machine Learning (ML). Abdominal planning CTs (n = 142) of recently-treated childhood cancer patients were gathered, their organs at risk were segmented, and 300 artificial Wilms' tumor plans were sampled automatically. Each artificial plan was automatically emulated on the 142 CTs, resulting in 42,600 3D dose distributions from which dose-volume metrics were derived. Anatomical features were extracted from digitally reconstructed radiographs simulated from the CTs to resemble historical radiographs. Further, patient and radiotherapy plan features typically available from historical treatment records were collected. An evolutionary ML algorithm was then used to link features to dosevolume metrics. Besides 5-fold cross validation, a further evaluation was done on an independent dataset of five CTs each associated with two clinical plans. Crossvalidation resulted in mean absolute errors ≤0.6 Gy for organs completely inside or outside the field. For organs positioned at the edge of the field, mean absolute errors ≤1.7 Gy for D mean , ≤2.9 Gy for D 2cc , and ≤13% for V 5Gy and V 10Gy , were obtained, without systematic bias. Similar results were found for the independent dataset. To conclude, we proposed a novel organ dose reconstruction method that uses ML models to predict dose-volume metric values given patient and plan features. Our approach is not only accurate, but also efficient, as the setup of a surrogate is no longer needed.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Automatic Generation Of Artificial Wilms' Tumor Plansmentioning

confidence: 99%

Section: Generation Of the Dataset For MLmentioning

confidence: 99%

Section: Explanatory Variables: Features Of Patients and Plansmentioning

confidence: 99%

Section: Explanatory Variables: Features Of Patients and Plansmentioning

confidence: 99%

See 3 more Smart Citations

Surrogate-free machine learning-based organ dose reconstruction for pediatric abdominal radiotherapy

Virgolin,

Wang,

Balgobind

et al. 2020

Preprint

View full text Add to dashboard Cite

show abstract

Surrogate-free machine learning-based organ dose reconstruction for pediatric abdominal radiotherapy

Virgolin

Wang²,

Balgobind³

et al. 2020

Phys. Med. Biol.

View full text Add to dashboard Cite

To study radiotherapy-related adverse effects, detailed dose information (3D distribution) is needed for accurate dose-effect modeling. For childhood cancer survivors who underwent radiotherapy in the pre-CT era, only 2D radiographs were acquired, thus 3D dose distributions must be reconstructed from limited information. State-of-the-art methods achieve this by using 3D surrogate anatomies. These can however lack personalization and lead to coarse reconstructions. We present and validate a surrogate-free dose reconstruction method based on Machine Learning (ML). Abdominal planning CTs (n = 142) of recently-treated childhood cancer patients were gathered, their organs at risk were segmented, and 300 artificial Wilms’ tumor plans were sampled automatically. Each artificial plan was automatically emulated on the 142 CTs, resulting in 42,600 3D dose distributions from which dose-volume metrics were derived. Anatomical features were extracted from digitally reconstructed radiographs simulated from the CTs to resemble historical radiographs. Further, patient and radiotherapy plan features typically available from historical treatment records were collected. An evolutionary ML algorithm was then used to link features to dose-volume metrics. Besides 5-fold cross validation, a further evaluation was done on an independent dataset of five CTs each associated with two clinical plans. Cross-validation resulted in mean absolute errors ≤ 0.6 Gy for organs completely inside or outside the field. For organs positioned at the edge of the field, mean absolute errors ≤ 1.7 Gy for D m e a n , ≤ 2.9 Gy for D 2 c c , and ≤ 13% for V 5 G y and V 10 G y , were obtained, without systematic bias. Similar results were found for the independent dataset. To conclude, we proposed a novel organ dose reconstruction method that uses ML models to predict dose-volume metric values given patient and plan features. Our approach is not only accurate, but also efficient, as the setup of a surrogate is no longer needed.

show abstract

Automatic generation of three-dimensional dose reconstruction data for two-dimensional radiotherapy plans for historically treated patients

Cited by 2 publications

References 46 publications

Surrogate-free machine learning-based organ dose reconstruction for pediatric abdominal radiotherapy

Surrogate-free machine learning-based organ dose reconstruction for pediatric abdominal radiotherapy

Surrogate-free machine learning-based organ dose reconstruction for pediatric abdominal radiotherapy

Contact Info

Product

Resources

About