Lin Ma
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Lin Ma has a total of 40 co-authors. Here are the top co-authors by number of shared publications:
- Weiguo Lu
University of Texas Southwestern Medical Center
7 shared publications - Xuejun Gu
Stanford University
5 shared publications Mingli Chen
Radiation Oncology, University of Texas Southwestern Medical Center at Dallas, 6363 Forest Park Rd, Dallas, Texas, 75390-9315, UNITED STATES
4 shared publications- Mingli Chen
Radiation Oncology, University of Texas Southwestern Medical Center at Dallas, 6363 Forest Park Rd, Dallas, Texas, 75390-9315, UNITED STATES
3 shared publications David R Grosshans
Radiation Oncology, The University of Texas MD Anderson Cancer Center Division of Radiation Oncology, 1515 Holcombe Blvd, Houston, Texas, 77030, UNITED STATES
2 shared publicationsDongxu Yang
Department of Radiation Oncology, The University of Texas Southwestern Medical Center, 6363 Forest Park Road, Dallas, Texas, 75390, UNITED STATES
2 shared publications
Latest Publications
- Open Access
- 4 March 2025
- Published by Elsevier in JACC: Cardiooncology
- Vol. 7 (3) , 219-230
Cardiac Effects of Modern Breast Radiation Therapy in Patients Receiving Systemic Cancer Therapy
- 4 March 2025
- Published by Elsevier in JACC: Cardiooncology
- Vol. 7 (3) , 219-230
Open AccessCardiac Effects of Modern Breast Radiation Therapy in Patients Receiving Systemic Cancer Therapy
- Open Access
- 17 July 2024
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 69 (15) , 155006
Investigation of intra-fractionated range guided adaptive proton therapy (RGAPT): II. Range-shift compensated on-line treatment adaptation and verification
We previously proposed range-guided adaptive proton therapy (RGAPT) that uses mid-range treatment beams as probing beams and intra-fractionated range measurements for online adaptation. In this work, we demonstrated experimental verification and reported the dosimetric accuracy for RGAPT. A STEEV phantom was used for the experiments, and a 3 × 3 × 3 cm3 cube inside the phantom was assigned to be the treatment target. We simulated three online range shift scenarios: reference, overshoot, and undershoot, by placing upstream Lucite sheets, 4, 0, and 8 that corresponded to changes of 0, 6.8, and −6.8 mm, respectively, in water-equivalent path length. The reference treatment plan was to deliver single-field uniform target doses in pencil beam scanning mode and generated on the Eclipse treatment planning system. Different numbers of mid-range layers, including single, three, and five layers, were selected as probing beams to evaluate beam range (BR) measurement accuracy in positron emission tomography (PET). Online plans were modified to adapt to BR shifts and compensate for probing beam doses. In contrast, non-adaptive plans were also delivered and compared to adaptive plans by film measurements. The mid-range probing beams of three (5.55MU) and five layers (8.71MU) yielded accurate range shift measurements in 60 s of PET acquisition with uncertainty of 0.5 mm while the single-layer probing (1.65MU) was not sufficient for measurements. The adaptive plans achieved an average gamma (2%/2 mm) passing rate of 95%. In contrast, the non-adaptive plans only had an average passing rate of 69%. RGAPT planning and delivery are feasible and verified by the experiments. The probing beam delivery, range measurements, and adaptive planning and delivery added a small increase in treatment delivery workflow time but resulted in substantial dose improvement. The three-layer mid-range probing was most suitable considering the balance of high range measurement accuracy and the low number of probing beam layers.
- 17 July 2024
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 69 (15) , 155006
Open AccessInvestigation of intra-fractionated range guided adaptive proton therapy (RGAPT): II. Range-shift compensated on-line treatment adaptation and verification
We previously proposed range-guided adaptive proton therapy (RGAPT) that uses mid-range treatment beams as probing beams and intra-fractionated range measurements for online adaptation. In this work, we demonstrated experimental verification and reported the dosimetric accuracy for RGAPT. A STEEV phantom was used for the experiments, and a 3 × 3 × 3 cm3 cube inside the phantom was assigned to be the treatment target. We simulated three online range shift scenarios: reference, overshoot, and undershoot, by placing upstream Lucite sheets, 4, 0, and 8 that corresponded to changes of 0, 6.8, and −6.8 mm, respectively, in water-equivalent path length. The reference treatment plan was to deliver single-field uniform target doses in pencil beam scanning mode and generated on the Eclipse treatment planning system. Different numbers of mid-range layers, including single, three, and five layers, were selected as probing beams to evaluate beam range (BR) measurement accuracy in positron emission tomography (PET). Online plans were modified to adapt to BR shifts and compensate for probing beam doses. In contrast, non-adaptive plans were also delivered and compared to adaptive plans by film measurements. The mid-range probing beams of three (5.55MU) and five layers (8.71MU) yielded accurate range shift measurements in 60 s of PET acquisition with uncertainty of 0.5 mm while the single-layer probing (1.65MU) was not sufficient for measurements. The adaptive plans achieved an average gamma (2%/2 mm) passing rate of 95%. In contrast, the non-adaptive plans only had an average passing rate of 69%. RGAPT planning and delivery are feasible and verified by the experiments. The probing beam delivery, range measurements, and adaptive planning and delivery added a small increase in treatment delivery workflow time but resulted in substantial dose improvement. The three-layer mid-range probing was most suitable considering the balance of high range measurement accuracy and the low number of probing beam layers.
- Open Access
- 17 July 2024
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 69 (15) , 155005
Investigation of intra-fractionated range guided adaptive proton therapy: I. On-line PET imaging and range measurement
Objective. Develop a prototype on-line positron emission tomography (PET) scanner and evaluate its capability of on-line imaging and intra-fractionated proton-induced radioactivity range measurement. Approach. Each detector consists of 32 × 32 array of 2 × 2 × 30 mm3 Lutetium–Yttrium Oxyorthosilicate scintillators with single-scintillator-end readout through a 20 × 20 array of 3 × 3 mm2 Silicon Photomultipliers. The PET can be configurated with a full-ring of 20 detectors for conventional PET imaging or a partial-ring of 18 detectors for on-line imaging and range measurement. All detector-level readout and processing electronics are attached to the backside of the system gantry and their output signals are transferred to a field-programable-gate-array based system electronics and data acquisition that can be placed 2 m away from the gantry. The PET imaging performance and radioactivity range measurement capability were evaluated by both the offline study that placed a radioactive source with known intensity and distribution within a phantom and the online study that irradiated a phantom with proton beams under different radiation and imaging conditions. Main results. The PET has 32 cm diameter and 6.5 cm axial length field-of-view (FOV), ∼2.3–5.0 mm spatial resolution within FOV, 3% sensitivity at the FOV center, 18%–30% energy resolution, and ∼9 ns coincidence time resolution. The offline study shows the PET can determine the shift of distal falloff edge position of a known radioactivity distribution with the accuracy of 0.3 ± 0.3 mm even without attenuation and scatter corrections, and online study shows the PET can measure the shift of proton-induced positron radioactive range with the accuracy of 0.6 ± 0.3 mm from the data acquired with a short-acquisition (60 s) and low-dose (5 MU) proton radiation to a human head phantom. Significance. This study demonstrated the capability of intra-fractionated PET imaging and radioactivity range measurement and will enable the investigation on the feasibility of intra-fractionated, range-shift compensated adaptive proton therapy.
- 17 July 2024
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 69 (15) , 155005
Open AccessInvestigation of intra-fractionated range guided adaptive proton therapy: I. On-line PET imaging and range measurement
Objective. Develop a prototype on-line positron emission tomography (PET) scanner and evaluate its capability of on-line imaging and intra-fractionated proton-induced radioactivity range measurement. Approach. Each detector consists of 32 × 32 array of 2 × 2 × 30 mm3 Lutetium–Yttrium Oxyorthosilicate scintillators with single-scintillator-end readout through a 20 × 20 array of 3 × 3 mm2 Silicon Photomultipliers. The PET can be configurated with a full-ring of 20 detectors for conventional PET imaging or a partial-ring of 18 detectors for on-line imaging and range measurement. All detector-level readout and processing electronics are attached to the backside of the system gantry and their output signals are transferred to a field-programable-gate-array based system electronics and data acquisition that can be placed 2 m away from the gantry. The PET imaging performance and radioactivity range measurement capability were evaluated by both the offline study that placed a radioactive source with known intensity and distribution within a phantom and the online study that irradiated a phantom with proton beams under different radiation and imaging conditions. Main results. The PET has 32 cm diameter and 6.5 cm axial length field-of-view (FOV), ∼2.3–5.0 mm spatial resolution within FOV, 3% sensitivity at the FOV center, 18%–30% energy resolution, and ∼9 ns coincidence time resolution. The offline study shows the PET can determine the shift of distal falloff edge position of a known radioactivity distribution with the accuracy of 0.3 ± 0.3 mm even without attenuation and scatter corrections, and online study shows the PET can measure the shift of proton-induced positron radioactive range with the accuracy of 0.6 ± 0.3 mm from the data acquired with a short-acquisition (60 s) and low-dose (5 MU) proton radiation to a human head phantom. Significance. This study demonstrated the capability of intra-fractionated PET imaging and radioactivity range measurement and will enable the investigation on the feasibility of intra-fractionated, range-shift compensated adaptive proton therapy.
- 6 December 2022
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 67 (24) , 245002
Ensemble learning for glioma patients overall survival prediction using pre-operative MRIs
Objective : Gliomas are the most common primary brain tumors. Approximately 70% of the glioma patients diagnosed with glioblastoma have an averaged overall survival (OS) of only ∼16 months. Early survival prediction is essential for treatment decision-making in glioma patients. Here we proposed an ensemble learning approach to predict the post-operative OS of glioma patients using only pre-operative MRIs. Approach : Our dataset was from the Medical Image Computing and Computer Assisted Intervention Brain Tumor Segmentation challenge 2020, which consists of multimodal pre-operative MRI scans of 235 glioma patients with survival days recorded. The backbone of our approach was a Siamese network consisting of twinned ResNet-based feature extractors followed by a 3-layer classifier. During training, the feature extractors explored traits of intra and inter-class by minimizing contrastive loss of randomly paired 2D pre-operative MRIs, and the classifier utilized the extracted features to generate labels with cost defined by cross-entropy loss. During testing, the extracted features were also utilized to define distance between the test sample and the reference composed of training data, to generate an additional predictor via K-NN classification. The final label was the ensemble classification from both the Siamese model and the K-NN model. Main results : Our approach classifies the glioma patients into 3 OS classes: long-survivors (>15 months), mid-survivors (between 10 and 15 months) and short-survivors (Significance : Our Siamese network based ensemble learning approach demonstrated promising ability in mining discriminative features with minimal manual processing and generalization requirement. This prediction strategy can be potentially applied to assist timely clinical decision-making.
- 6 December 2022
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 67 (24) , 245002
Ensemble learning for glioma patients overall survival prediction using pre-operative MRIs
Objective : Gliomas are the most common primary brain tumors. Approximately 70% of the glioma patients diagnosed with glioblastoma have an averaged overall survival (OS) of only ∼16 months. Early survival prediction is essential for treatment decision-making in glioma patients. Here we proposed an ensemble learning approach to predict the post-operative OS of glioma patients using only pre-operative MRIs. Approach : Our dataset was from the Medical Image Computing and Computer Assisted Intervention Brain Tumor Segmentation challenge 2020, which consists of multimodal pre-operative MRI scans of 235 glioma patients with survival days recorded. The backbone of our approach was a Siamese network consisting of twinned ResNet-based feature extractors followed by a 3-layer classifier. During training, the feature extractors explored traits of intra and inter-class by minimizing contrastive loss of randomly paired 2D pre-operative MRIs, and the classifier utilized the extracted features to generate labels with cost defined by cross-entropy loss. During testing, the extracted features were also utilized to define distance between the test sample and the reference composed of training data, to generate an additional predictor via K-NN classification. The final label was the ensemble classification from both the Siamese model and the K-NN model. Main results : Our approach classifies the glioma patients into 3 OS classes: long-survivors (>15 months), mid-survivors (between 10 and 15 months) and short-survivors (Significance : Our Siamese network based ensemble learning approach demonstrated promising ability in mining discriminative features with minimal manual processing and generalization requirement. This prediction strategy can be potentially applied to assist timely clinical decision-making.
- 19 February 2022
- Published by Wiley in Medical Physics
- Vol. 49 (4) , 2159-2171
Volumetric dose extension for isodose tuning
Purpose To develop a method that can extend dose from two isodose surfaces (isosurfaces) to the entire patient volume, and to demonstrate its application in radiotherapy plan isodose tuning. Methods We hypothesized that volumetric dose distribution can be extended from two isosurfaces—the 100% isosurface and a reference isosurface—-with the distances to these two surfaces (Lref) as extension variables. The extension function is modeled by a three-dimensional lookup table (LUT), where voxel dose values from clinical plans are binned by three indexes: L100, Lref and Dref (reference dose level). The mean and standard deviation of voxel doses in each bin are calculated and stored in LUT. Volumetric dose extension is performed voxelwisely by indexing the LUT with the L100, Lref and Dref of each query voxel. The mean dose stored in the corresponding bin is filled into the query voxel as extended dose, and the standard deviation be filled voxelwisely as the uncertainty of extension result. We applied dose extension in isodose tuning, which aims to tune volumetric dose distribution by isosurface dragging. We adopted extended dose as an approximate dose estimation, and combined it with dose correction strategy to achieve accurate dose tuning. Results We collected 32 post-operative prostate volumetric modulated arc therapy (VMAT) cases and built the LUT and its associated uncertainties from the doses of 27 cases. The dose extension method was tested on five cases, whose dose distributions were defined as ground truth (GT). We extended the doses from 100% and 50% GT isosurfaces to the entire volume, and evaluated the accuracy of extended doses. The 5 mm/5% gamma passing rate (GPR) of extended doses are 92.0%. The mean error is 4.5%, which is consistent to the uncertainty estimated by LUT. The dose difference in 90.5% of voxels is within two sigma and 97.5% in three sigma. The calculation time is less than two seconds. To simulate plan isodose tuning, we optimized a dose with less sparing on rectum (than GT dose) and defined it as a “base dose”—the dose awaiting isosurface dragging. In front-end, the simulated isodose tuning is conducted as such that the base dose was given to plan tuner, and its 50% isosurface would be dragged to the desired position (position of 50% isosurface in GT dose). In back-end, the output of isodose tuning is obtained by 1) extending dose from the desired isosurfaces and viewed the extended dose as an approximate dose, 2) obtaining a correction map from the base dose, and 3) applying the correction map to the extended dose. The accuracy of output—extended dose with correction—was 97.2% in GPR (3 mm/3%) and less than 1% in mean dose difference. The total calculation time is less than two seconds, which allows for interactive isodose tuning. Conclusions We developed a dose extension method that generates volumetric dose distribution from two surfaces. The application of dose extension is in interactive isodose tuning. The distance-based LUT fashion and correction strategy guarantee the computation efficiency and accuracy in isodose tuning. This article is protected by copyright. All rights reserved
- 19 February 2022
- Published by Wiley in Medical Physics
- Vol. 49 (4) , 2159-2171
Volumetric dose extension for isodose tuning
Purpose To develop a method that can extend dose from two isodose surfaces (isosurfaces) to the entire patient volume, and to demonstrate its application in radiotherapy plan isodose tuning. Methods We hypothesized that volumetric dose distribution can be extended from two isosurfaces—the 100% isosurface and a reference isosurface—-with the distances to these two surfaces (Lref) as extension variables. The extension function is modeled by a three-dimensional lookup table (LUT), where voxel dose values from clinical plans are binned by three indexes: L100, Lref and Dref (reference dose level). The mean and standard deviation of voxel doses in each bin are calculated and stored in LUT. Volumetric dose extension is performed voxelwisely by indexing the LUT with the L100, Lref and Dref of each query voxel. The mean dose stored in the corresponding bin is filled into the query voxel as extended dose, and the standard deviation be filled voxelwisely as the uncertainty of extension result. We applied dose extension in isodose tuning, which aims to tune volumetric dose distribution by isosurface dragging. We adopted extended dose as an approximate dose estimation, and combined it with dose correction strategy to achieve accurate dose tuning. Results We collected 32 post-operative prostate volumetric modulated arc therapy (VMAT) cases and built the LUT and its associated uncertainties from the doses of 27 cases. The dose extension method was tested on five cases, whose dose distributions were defined as ground truth (GT). We extended the doses from 100% and 50% GT isosurfaces to the entire volume, and evaluated the accuracy of extended doses. The 5 mm/5% gamma passing rate (GPR) of extended doses are 92.0%. The mean error is 4.5%, which is consistent to the uncertainty estimated by LUT. The dose difference in 90.5% of voxels is within two sigma and 97.5% in three sigma. The calculation time is less than two seconds. To simulate plan isodose tuning, we optimized a dose with less sparing on rectum (than GT dose) and defined it as a “base dose”—the dose awaiting isosurface dragging. In front-end, the simulated isodose tuning is conducted as such that the base dose was given to plan tuner, and its 50% isosurface would be dragged to the desired position (position of 50% isosurface in GT dose). In back-end, the output of isodose tuning is obtained by 1) extending dose from the desired isosurfaces and viewed the extended dose as an approximate dose, 2) obtaining a correction map from the base dose, and 3) applying the correction map to the extended dose. The accuracy of output—extended dose with correction—was 97.2% in GPR (3 mm/3%) and less than 1% in mean dose difference. The total calculation time is less than two seconds, which allows for interactive isodose tuning. Conclusions We developed a dose extension method that generates volumetric dose distribution from two surfaces. The application of dose extension is in interactive isodose tuning. The distance-based LUT fashion and correction strategy guarantee the computation efficiency and accuracy in isodose tuning. This article is protected by copyright. All rights reserved
- 19 January 2022
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 67 (2) , 025004
Deep-learning and radiomics ensemble classifier for false positive reduction in brain metastases segmentation
Stereotactic radiosurgery (SRS) is now the standard of care for brain metastases (BMs) patients. The SRS treatment planning process requires precise target delineation, which in clinical workflow for patients with multiple (>4) BMs (mBMs) could become a pronounced time bottleneck. Our group has developed an automated BMs segmentation platform to assist in this process. The accuracy of the auto-segmentation, however, is influenced by the presence of false-positive segmentations, mainly caused by the injected contrast during MRI acquisition. To address this problem and further improve the segmentation performance, a deep-learning and radiomics ensemble classifier was developed to reduce the false-positive rate in segmentations. The proposed model consists of a Siamese network and a radiomic-based support vector machine (SVM) classifier. The 2D-based Siamese network contains a pair of parallel feature extractors with shared weights followed by a single classifier. This architecture is designed to identify the inter-class difference. On the other hand, the SVM model takes the radiomic features extracted from 3D segmentation volumes as the input for twofold classification, either a false-positive segmentation or a true BM. Lastly, the outputs from both models create an ensemble to generate the final label. The performance of the proposed model in the segmented mBMs testing dataset reached the accuracy (ACC), sensitivity (SEN), specificity (SPE) and area under the curve of 0.91, 0.96, 0.90 and 0.93, respectively. After integrating the proposed model into the original segmentation platform, the average segmentation false negative rate (FNR) and the false positive over the union (FPoU) were 0.13 and 0.09, respectively, which preserved the initial FNR (0.07) and significantly improved the FPoU (0.55). The proposed method effectively reduced the false-positive rate in the BMs raw segmentations indicating that the integration of the proposed ensemble classifier into the BMs segmentation platform provides a beneficial tool for mBMs SRS management.
- 19 January 2022
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 67 (2) , 025004
Deep-learning and radiomics ensemble classifier for false positive reduction in brain metastases segmentation
Stereotactic radiosurgery (SRS) is now the standard of care for brain metastases (BMs) patients. The SRS treatment planning process requires precise target delineation, which in clinical workflow for patients with multiple (>4) BMs (mBMs) could become a pronounced time bottleneck. Our group has developed an automated BMs segmentation platform to assist in this process. The accuracy of the auto-segmentation, however, is influenced by the presence of false-positive segmentations, mainly caused by the injected contrast during MRI acquisition. To address this problem and further improve the segmentation performance, a deep-learning and radiomics ensemble classifier was developed to reduce the false-positive rate in segmentations. The proposed model consists of a Siamese network and a radiomic-based support vector machine (SVM) classifier. The 2D-based Siamese network contains a pair of parallel feature extractors with shared weights followed by a single classifier. This architecture is designed to identify the inter-class difference. On the other hand, the SVM model takes the radiomic features extracted from 3D segmentation volumes as the input for twofold classification, either a false-positive segmentation or a true BM. Lastly, the outputs from both models create an ensemble to generate the final label. The performance of the proposed model in the segmented mBMs testing dataset reached the accuracy (ACC), sensitivity (SEN), specificity (SPE) and area under the curve of 0.91, 0.96, 0.90 and 0.93, respectively. After integrating the proposed model into the original segmentation platform, the average segmentation false negative rate (FNR) and the false positive over the union (FPoU) were 0.13 and 0.09, respectively, which preserved the initial FNR (0.07) and significantly improved the FPoU (0.55). The proposed method effectively reduced the false-positive rate in the BMs raw segmentations indicating that the integration of the proposed ensemble classifier into the BMs segmentation platform provides a beneficial tool for mBMs SRS management.
- 27 November 2020
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 65 (23) , 235035
Deep learning-based inverse mapping for fluence map prediction
We developed a fluence map prediction method that directly generates fluence maps for a given desired dose distribution without optimization for volumetric modulated arc therapy (VMAT) planning. The prediction consists of two steps. First, projections of the desired dose are calculated and then inversely mapped to fluence maps in the phantom geometry by a deep neural network. Second, a plan scaling technique is applied to scale fluence maps from phantom to patient geometry. We evaluated the performance of the proposed fluence map prediction method for 102 head and neck (H&N) and 14 prostate cancer VMAT plans by comparing the patient doses calculated from the predicted fluence maps with the given desired dose distributions. The mean dose differences were 1.42% ± 0.37%, 1.53% ± 0.44% and 1.25% ± 0.44% for the planning target volume (PTV), the region from the PTV boundary to the 50% isodose line, and the region from the 50% to the 20% isodose line, respectively. The gamma passing rate was 98.06% ± 2.64% with the 3 mm/3% criterion. The prediction time for a single VMAT plan was less than one second. In conclusion, we developed an inverse mapping-based method that predicts fluence maps for desired dose distributions with high accuracy. Our method is effectively an optimization-free inverse planning approach, which was orders of magnitude faster than fluence map optimization. Combining the proposed method with leaf sequencing has the potential to dramatically speed up VMAT treatment planning.
- 27 November 2020
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 65 (23) , 235035
Deep learning-based inverse mapping for fluence map prediction
We developed a fluence map prediction method that directly generates fluence maps for a given desired dose distribution without optimization for volumetric modulated arc therapy (VMAT) planning. The prediction consists of two steps. First, projections of the desired dose are calculated and then inversely mapped to fluence maps in the phantom geometry by a deep neural network. Second, a plan scaling technique is applied to scale fluence maps from phantom to patient geometry. We evaluated the performance of the proposed fluence map prediction method for 102 head and neck (H&N) and 14 prostate cancer VMAT plans by comparing the patient doses calculated from the predicted fluence maps with the given desired dose distributions. The mean dose differences were 1.42% ± 0.37%, 1.53% ± 0.44% and 1.25% ± 0.44% for the planning target volume (PTV), the region from the PTV boundary to the 50% isodose line, and the region from the 50% to the 20% isodose line, respectively. The gamma passing rate was 98.06% ± 2.64% with the 3 mm/3% criterion. The prediction time for a single VMAT plan was less than one second. In conclusion, we developed an inverse mapping-based method that predicts fluence maps for desired dose distributions with high accuracy. Our method is effectively an optimization-free inverse planning approach, which was orders of magnitude faster than fluence map optimization. Combining the proposed method with leaf sequencing has the potential to dramatically speed up VMAT treatment planning.
- 20 November 2020
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 65 (23) , 235001
Deep learning-based medical image segmentation with limited labels
Deep learning (DL)-based auto-segmentation has the potential for accurate organ delineation in radiotherapy applications but requires large amounts of clean labeled data to train a robust model. However, annotating medical images is extremely time-consuming and requires clinical expertise, especially for segmentation that demands voxel-wise labels. On the other hand, medical images without annotations are abundant and highly accessible. To alleviate the influence of the limited number of clean labels, we propose a weakly supervised DL training approach using deformable image registration (DIR)-based annotations, leveraging the abundance of unlabeled data. We generate pseudo-contours by utilizing DIR to propagate atlas contours onto abundant unlabeled images and train a robust DL-based segmentation model. With 10 labeled TCIA dataset and 50 unlabeled CT scans from our institution, our model achieved Dice similarity coefficient of 87.9%, 73.4%, 73.4%, 63.2% and 61.0% on mandible, left & right parotid glands and left & right submandibular glands of TCIA test set and competitive performance on our institutional clinical dataset and a third party (PDDCA) dataset. Experimental results demonstrated the proposed method outperformed traditional multi-atlas DIR methods and fully supervised limited data training and is promising for DL-based medical image segmentation application with limited annotated data.
- 20 November 2020
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 65 (23) , 235001
Deep learning-based medical image segmentation with limited labels
Deep learning (DL)-based auto-segmentation has the potential for accurate organ delineation in radiotherapy applications but requires large amounts of clean labeled data to train a robust model. However, annotating medical images is extremely time-consuming and requires clinical expertise, especially for segmentation that demands voxel-wise labels. On the other hand, medical images without annotations are abundant and highly accessible. To alleviate the influence of the limited number of clean labels, we propose a weakly supervised DL training approach using deformable image registration (DIR)-based annotations, leveraging the abundance of unlabeled data. We generate pseudo-contours by utilizing DIR to propagate atlas contours onto abundant unlabeled images and train a robust DL-based segmentation model. With 10 labeled TCIA dataset and 50 unlabeled CT scans from our institution, our model achieved Dice similarity coefficient of 87.9%, 73.4%, 73.4%, 63.2% and 61.0% on mandible, left & right parotid glands and left & right submandibular glands of TCIA test set and competitive performance on our institutional clinical dataset and a third party (PDDCA) dataset. Experimental results demonstrated the proposed method outperformed traditional multi-atlas DIR methods and fully supervised limited data training and is promising for DL-based medical image segmentation application with limited annotated data.