Gayoung Kim
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Research Interests
Gayoung Kim has a total of 26 co-authors. Here are the top co-authors by number of shared publications:
- Junghoon Lee
Department of Radiation Oncology and Molecular Radiation Sciences, Johns Hopkins University, Baltimore, MD
4 shared publications - Michael Roumeliotis
Department of Radiation Oncology and Molecular Radiation Sciences, The Johns Hopkins University School of Medicine, Baltimore, MD
3 shared publications - Sung Min Kim
Department of Medical Device and Healthcare, Dongguk University-Seoul, Seoul, Republic of Korea
3 shared publications Akila N. Viswanathan
Department of Radiation Oncology and Molecular Radiation Sciences, The Johns Hopkins University School of Medicine, Baltimore, MD
2 shared publications- Ehud J. Schmidt
Department of Radiation Oncology and Molecular Radiation Sciences, The Johns Hopkins University School of Medicine, Baltimore, MD; Department of Medicine (Cardiology), The Johns Hopkins University School of Medicine, Baltimore, MD
2 shared publications Majd Antaki
Physics and Advanced Development, Elekta Inc., Washington, DC
2 shared publications
Latest Publications
- 7 April 2026
- Published by Elsevier in Brachytherapy
- Vol. 25 (5) , 741-748
Prospective operational feasibility, safety, and workflow of magnetic resonance-guided tracking technology for interstitial gynecologic brachytherapy
- 7 April 2026
- Published by Elsevier in Brachytherapy
- Vol. 25 (5) , 741-748
Prospective operational feasibility, safety, and workflow of magnetic resonance-guided tracking technology for interstitial gynecologic brachytherapy
- 7 November 2025
- Published by Elsevier in Brachytherapy
- Vol. 25 (1) , 85-94
Automatic digitization of applicator and catheters for MRI-guided cervical cancer brachytherapy
- 7 November 2025
- Published by Elsevier in Brachytherapy
- Vol. 25 (1) , 85-94
Automatic digitization of applicator and catheters for MRI-guided cervical cancer brachytherapy
- 18 October 2024
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 69 (21) , 215014
Dual convolution-transformer UNet (DCT-UNet) for organs at risk and clinical target volume segmentation in MRI for cervical cancer brachytherapy
Objective. MRI is the standard imaging modality for high-dose-rate brachytherapy of cervical cancer. Precise contouring of organs at risk (OARs) and high-risk clinical target volume (HR-CTV) from MRI is a crucial step for radiotherapy planning and treatment. However, conventional manual contouring has limitations in terms of accuracy as well as procedural time. To overcome these, we propose a deep learning approach to automatically segment OARs (bladder, rectum, and sigmoid colon) and HR-CTV from female pelvic MRI. Approach. In the proposed pipeline, a coarse multi-organ segmentation model first segments all structures, from which a region of interest is computed for each structure. Then, each organ is segmented using an organ-specific fine segmentation model separately trained for each organ. To account for variable sizes of HR-CTV, a size-adaptive multi-model approach was employed. For coarse and fine segmentations, we designed a dual convolution-transformer UNet (DCT-UNet) which uses dual-path encoder consisting of convolution and transformer blocks. To evaluate our model, OAR segmentations were compared to the clinical contours drawn by the attending radiation oncologist. For HR-CTV, four sets of contours (clinical + three additional sets) were obtained to produce a consensus ground truth as well as for inter/intra-observer variability analysis. Main results. DCT-UNet achieved dice similarity coefficient (mean ± SD) of 0.932 ± 0.032 (bladder), 0.786 ± 0.090 (rectum), 0.663 ± 0.180 (sigmoid colon), and 0.741 ± 0.076 (HR-CTV), outperforming other state-of-the-art models. Notably, the size-adaptive multi-model significantly improved HR-CTV segmentation compared to a single-model. Furthermore, significant inter/intra-observer variability was observed, and our model showed comparable performance to all observers. Computation time for the entire pipeline per subject was 12.59 ± 0.79 s, which is significantly shorter than the typical manual contouring time of >15 min. Significance. These experimental results demonstrate that our model has great utility in cervical cancer brachytherapy by enabling fast and accurate automatic segmentation, and has potential in improving consistency in contouring. DCT-UNet source code is available at https://github.com/JHU-MICA/DCT-UNet.
- 18 October 2024
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 69 (21) , 215014
Dual convolution-transformer UNet (DCT-UNet) for organs at risk and clinical target volume segmentation in MRI for cervical cancer brachytherapy
Objective. MRI is the standard imaging modality for high-dose-rate brachytherapy of cervical cancer. Precise contouring of organs at risk (OARs) and high-risk clinical target volume (HR-CTV) from MRI is a crucial step for radiotherapy planning and treatment. However, conventional manual contouring has limitations in terms of accuracy as well as procedural time. To overcome these, we propose a deep learning approach to automatically segment OARs (bladder, rectum, and sigmoid colon) and HR-CTV from female pelvic MRI. Approach. In the proposed pipeline, a coarse multi-organ segmentation model first segments all structures, from which a region of interest is computed for each structure. Then, each organ is segmented using an organ-specific fine segmentation model separately trained for each organ. To account for variable sizes of HR-CTV, a size-adaptive multi-model approach was employed. For coarse and fine segmentations, we designed a dual convolution-transformer UNet (DCT-UNet) which uses dual-path encoder consisting of convolution and transformer blocks. To evaluate our model, OAR segmentations were compared to the clinical contours drawn by the attending radiation oncologist. For HR-CTV, four sets of contours (clinical + three additional sets) were obtained to produce a consensus ground truth as well as for inter/intra-observer variability analysis. Main results. DCT-UNet achieved dice similarity coefficient (mean ± SD) of 0.932 ± 0.032 (bladder), 0.786 ± 0.090 (rectum), 0.663 ± 0.180 (sigmoid colon), and 0.741 ± 0.076 (HR-CTV), outperforming other state-of-the-art models. Notably, the size-adaptive multi-model significantly improved HR-CTV segmentation compared to a single-model. Furthermore, significant inter/intra-observer variability was observed, and our model showed comparable performance to all observers. Computation time for the entire pipeline per subject was 12.59 ± 0.79 s, which is significantly shorter than the typical manual contouring time of >15 min. Significance. These experimental results demonstrate that our model has great utility in cervical cancer brachytherapy by enabling fast and accurate automatic segmentation, and has potential in improving consistency in contouring. DCT-UNet source code is available at https://github.com/JHU-MICA/DCT-UNet.
- 1 March 2024
- Published by SPIE-Intl Soc Optical Eng in Journal of Medical Imaging
- Vol. 11 (02) , 024009-17
Cascaded cross-attention transformers and convolutional neural networks for multi-organ segmentation in male pelvic computed tomography
PurposeSegmentation of the prostate and surrounding organs at risk from computed tomography is required for radiation therapy treatment planning. We propose an automatic two-step deep learning-based segmentation pipeline that consists of an initial multi-organ segmentation network for organ localization followed by organ-specific fine segmentation.ApproachInitial segmentation of all target organs is performed using a hybrid convolutional-transformer model, axial cross-attention UNet. The output from this model allows for region of interest computation and is used to crop tightly around individual organs for organ-specific fine segmentation. Information from this network is also propagated to the fine segmentation stage through an image enhancement module, highlighting regions of interest in the original image that might be difficult to segment. Organ-specific fine segmentation is performed on these cropped and enhanced images to produce the final output segmentation.ResultsWe apply the proposed approach to segment the prostate, bladder, rectum, seminal vesicles, and femoral heads from male pelvic computed tomography (CT). When tested on a held-out test set of 30 images, our two-step pipeline outperformed other deep learning-based multi-organ segmentation algorithms, achieving average dice similarity coefficient (DSC) of 0.836±0.071 (prostate), 0.947±0.038 (bladder), 0.828±0.057 (rectum), 0.724±0.101 (seminal vesicles), and 0.933±0.020 (femoral heads).ConclusionsOur results demonstrate that a two-step segmentation pipeline with initial multi-organ segmentation and additional fine segmentation can delineate male pelvic CT organs well. The utility of this additional layer of fine segmentation is most noticeable in challenging cases, as our two-step pipeline produces noticeably more accurate and less erroneous results compared to other state-of-the-art methods on such images.
- 1 March 2024
- Published by SPIE-Intl Soc Optical Eng in Journal of Medical Imaging
- Vol. 11 (02) , 024009-17
Cascaded cross-attention transformers and convolutional neural networks for multi-organ segmentation in male pelvic computed tomography
PurposeSegmentation of the prostate and surrounding organs at risk from computed tomography is required for radiation therapy treatment planning. We propose an automatic two-step deep learning-based segmentation pipeline that consists of an initial multi-organ segmentation network for organ localization followed by organ-specific fine segmentation.ApproachInitial segmentation of all target organs is performed using a hybrid convolutional-transformer model, axial cross-attention UNet. The output from this model allows for region of interest computation and is used to crop tightly around individual organs for organ-specific fine segmentation. Information from this network is also propagated to the fine segmentation stage through an image enhancement module, highlighting regions of interest in the original image that might be difficult to segment. Organ-specific fine segmentation is performed on these cropped and enhanced images to produce the final output segmentation.ResultsWe apply the proposed approach to segment the prostate, bladder, rectum, seminal vesicles, and femoral heads from male pelvic computed tomography (CT). When tested on a held-out test set of 30 images, our two-step pipeline outperformed other deep learning-based multi-organ segmentation algorithms, achieving average dice similarity coefficient (DSC) of 0.836±0.071 (prostate), 0.947±0.038 (bladder), 0.828±0.057 (rectum), 0.724±0.101 (seminal vesicles), and 0.933±0.020 (femoral heads).ConclusionsOur results demonstrate that a two-step segmentation pipeline with initial multi-organ segmentation and additional fine segmentation can delineate male pelvic CT organs well. The utility of this additional layer of fine segmentation is most noticeable in challenging cases, as our two-step pipeline produces noticeably more accurate and less erroneous results compared to other state-of-the-art methods on such images.
- Open Access
- 30 July 2022
- Published by Wiley in BioMed Research International
- Vol. 2022 (1) , 1705338
Robust Detection Model of Vascular Landmarks for Retinal Image Registration: A Two‐Stage Convolutional Neural Network
Registration is useful for image processing in computer vision. It can be applied to retinal images and provide support for ophthalmologists in tracking disease progression and monitoring therapeutic responses. This study proposed a robust detection model of vascular landmarks to improve the performance of retinal image registration. The proposed model consists of a two‐stage convolutional neural network, in which one segments the retinal vessels on a pair of images, and the other detects junction points from the vessel segmentation image. Information obtained from the model was utilized for the registration. The keypoints were extracted based on the acquired vascular landmark points, and the orientation features were calculated as descriptors. Then, the reference and sensed images were registered by matching keypoints using a homography matrix and random sample consensus algorithm. The proposed method was evaluated on five databases and seven evaluation metrics to verify both clinical effectiveness and robustness. The results established that the proposed method showed outstanding performance for registration compared with other state‐of‐the‐art methods. In particular, the high and significantly improved registration results were identified on FIRE database with area under the curve (AUC) of 0.988, 0.511, and 0.803 in S, P, and A classes. Furthermore, the proposed method worked well on poor quality and multimodal datasets demonstrating an ability to achieve high AUC above 0.8.
- 30 July 2022
- Published by Wiley in BioMed Research International
- Vol. 2022 (1) , 1705338
Open AccessRobust Detection Model of Vascular Landmarks for Retinal Image Registration: A Two‐Stage Convolutional Neural Network
Registration is useful for image processing in computer vision. It can be applied to retinal images and provide support for ophthalmologists in tracking disease progression and monitoring therapeutic responses. This study proposed a robust detection model of vascular landmarks to improve the performance of retinal image registration. The proposed model consists of a two‐stage convolutional neural network, in which one segments the retinal vessels on a pair of images, and the other detects junction points from the vessel segmentation image. Information obtained from the model was utilized for the registration. The keypoints were extracted based on the acquired vascular landmark points, and the orientation features were calculated as descriptors. Then, the reference and sensed images were registered by matching keypoints using a homography matrix and random sample consensus algorithm. The proposed method was evaluated on five databases and seven evaluation metrics to verify both clinical effectiveness and robustness. The results established that the proposed method showed outstanding performance for registration compared with other state‐of‐the‐art methods. In particular, the high and significantly improved registration results were identified on FIRE database with area under the curve (AUC) of 0.988, 0.511, and 0.803 in S, P, and A classes. Furthermore, the proposed method worked well on poor quality and multimodal datasets demonstrating an ability to achieve high AUC above 0.8.
- Open Access
- 30 December 2019
- Published by Springer Nature in BMC Medical Imaging
- Vol. 19 (1) , 1-13
Automated classification of dense calcium tissues in gray-scale intravascular ultrasound images using a deep belief network
Background: IVUS is widely used to quantitatively assess coronary artery disease. The purpose of this study was to automatically characterize dense calcium (DC) tissue in the gray scale intravascular ultrasound (IVUS) images using the image textural features. Methods: A total of 316 Gy-scale IVUS and corresponding virtual histology images from 26 patients with acute coronary syndrome who underwent IVUS along with X-ray angiography between October 2009 to September 2014 were retrospectively acquired and analyzed. One expert performed all procedures and assessed their IVUS scans. After image acquisition, the DC candidate and corresponding acoustic shadow regions were automatically determined. Then, nine image-base feature groups were extracted from the DC candidates. In order to reduce the dimensionalities, principal component analysis (PCA) was performed, and selected feature sets were utilized as an input for a deep belief network. Classification results were validated using 10-fold cross validation. Results: The dimensionality of the feature map was efficiently reduced by 50% (from 66 to 33) without any performance decrease using PCA method. Sensitivity, specificity, and accuracy of the proposed method were 92.8 ± 0.1%, 85.1 ± 0.1%, and 88.4 ± 0.1%, respectively (p < 0.05). We found that the window size could largely influence the characterization results, and selected the 5 × 5 size as the best condition. We also validated the performance superiority of the proposed method with traditional classification methods. Conclusions: These experimental results suggest that the proposed method has significant clinical applicability for IVUS-based cardiovascular diagnosis.
- 30 December 2019
- Published by Springer Nature in BMC Medical Imaging
- Vol. 19 (1) , 1-13
Open AccessAutomated classification of dense calcium tissues in gray-scale intravascular ultrasound images using a deep belief network
Background: IVUS is widely used to quantitatively assess coronary artery disease. The purpose of this study was to automatically characterize dense calcium (DC) tissue in the gray scale intravascular ultrasound (IVUS) images using the image textural features. Methods: A total of 316 Gy-scale IVUS and corresponding virtual histology images from 26 patients with acute coronary syndrome who underwent IVUS along with X-ray angiography between October 2009 to September 2014 were retrospectively acquired and analyzed. One expert performed all procedures and assessed their IVUS scans. After image acquisition, the DC candidate and corresponding acoustic shadow regions were automatically determined. Then, nine image-base feature groups were extracted from the DC candidates. In order to reduce the dimensionalities, principal component analysis (PCA) was performed, and selected feature sets were utilized as an input for a deep belief network. Classification results were validated using 10-fold cross validation. Results: The dimensionality of the feature map was efficiently reduced by 50% (from 66 to 33) without any performance decrease using PCA method. Sensitivity, specificity, and accuracy of the proposed method were 92.8 ± 0.1%, 85.1 ± 0.1%, and 88.4 ± 0.1%, respectively (p < 0.05). We found that the window size could largely influence the characterization results, and selected the 5 × 5 size as the best condition. We also validated the performance superiority of the proposed method with traditional classification methods. Conclusions: These experimental results suggest that the proposed method has significant clinical applicability for IVUS-based cardiovascular diagnosis.
- Open Access
- 6 November 2018
- Published by Springer Nature in BioMedical Engineering OnLine
- Vol. 17 (S2) , 1-15
A novel intensity-based multi-level classification approach for coronary plaque characterization in intravascular ultrasound images
Intravascular ultrasound (IVUS) is a commonly used diagnostic imaging method for coronary artery disease. Virtual histology (VH) characterizes the plaque components into fibrous tissue (FT), fibro-fatty tissue (FFT), necrotic core (NC), or dense calcium (DC). However, VH can obtain only a single-frame image in one cardiac cycle, and specific software is needed to obtain the radio frequency data. This study proposed a novel intensity-based multi-level classification model for plaque characterization. The plaque-containing regions between the intima and the media-adventitia were segmented manually for all IVUS frames. A total of 54 features including first order statistics, grey level co-occurrence matrix, Law’s energy measures, extended grey level run length matrix, intensity, and local binary pattern were estimated from the plaque-containing regions. After feature extraction, optimal features were selected using principle component analysis (PCA), and these were utilized as the input for the classification models. Plaque components were classified into FT, FFT, NC, or DC using an intensity-based multi-level classification model consisting of three different nets. Net 1 differentiated low-intensity components into FT/FFT and NC/DC groups. Then, net 2 subsequently divided FT/FFT into FT or FFT, whereas the remainder and high-intensity components were classified into NC or DC via net 3. To improve classification accuracy, each net utilized three different input features obtained by PCA. Classification performance was evaluated in terms of sensitivity, specificity, accuracy, and receiver operating characteristic curve. Quantitative results indicated that the proposed method showed significantly high classification accuracy for all tissue types. The classifiers had classification accuracies of 85.1%, 71.9%, and 77.2%, respectively, and the areas under the curve were 0.845, 0.704, and 0.783. In particular, the proposed method achieved relatively high sensitivity (82.0%) and specificity (87.1%) for differentiating between the FT/FFT and NC/DC groups. These results confirmed the clinical applicability of the proposed approach for IVUS-based tissue characterization.
- 6 November 2018
- Published by Springer Nature in BioMedical Engineering OnLine
- Vol. 17 (S2) , 1-15
Open AccessA novel intensity-based multi-level classification approach for coronary plaque characterization in intravascular ultrasound images
Intravascular ultrasound (IVUS) is a commonly used diagnostic imaging method for coronary artery disease. Virtual histology (VH) characterizes the plaque components into fibrous tissue (FT), fibro-fatty tissue (FFT), necrotic core (NC), or dense calcium (DC). However, VH can obtain only a single-frame image in one cardiac cycle, and specific software is needed to obtain the radio frequency data. This study proposed a novel intensity-based multi-level classification model for plaque characterization. The plaque-containing regions between the intima and the media-adventitia were segmented manually for all IVUS frames. A total of 54 features including first order statistics, grey level co-occurrence matrix, Law’s energy measures, extended grey level run length matrix, intensity, and local binary pattern were estimated from the plaque-containing regions. After feature extraction, optimal features were selected using principle component analysis (PCA), and these were utilized as the input for the classification models. Plaque components were classified into FT, FFT, NC, or DC using an intensity-based multi-level classification model consisting of three different nets. Net 1 differentiated low-intensity components into FT/FFT and NC/DC groups. Then, net 2 subsequently divided FT/FFT into FT or FFT, whereas the remainder and high-intensity components were classified into NC or DC via net 3. To improve classification accuracy, each net utilized three different input features obtained by PCA. Classification performance was evaluated in terms of sensitivity, specificity, accuracy, and receiver operating characteristic curve. Quantitative results indicated that the proposed method showed significantly high classification accuracy for all tissue types. The classifiers had classification accuracies of 85.1%, 71.9%, and 77.2%, respectively, and the areas under the curve were 0.845, 0.704, and 0.783. In particular, the proposed method achieved relatively high sensitivity (82.0%) and specificity (87.1%) for differentiating between the FT/FFT and NC/DC groups. These results confirmed the clinical applicability of the proposed approach for IVUS-based tissue characterization.