Theodoros Panagiotis Vagenas
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Theodoros Panagiotis Vagenas has a total of 22 co-authors. Here are the top co-authors by number of shared publications:
- George K. Matsopoulos
Biomedical Engineering Laboratory, Institute of Communication and Computer Systems of the National Technical University of Athens, 15773 Athens, Greece
10 shared publications - Ioannis Kakkos
Biomedical Engineering Laboratory, School of Electrical & Computer Engineering, National Technical University of Athens, 15773 Athens, Greece
4 shared publications - Konstantinos Georgas
National Technical University of Athens
4 shared publications - Ioannis Vezakis
Biomedical Engineering Laboratory, School of Electrical & Computer Engineering, National Technical University of Athens, 15773 Athens, Greece
3 shared publications Antonia Dimitrakopoulou-Strauss
Clinical Cooperation Unit Nuclear Medicine, German Cancer Research Center, Heidelberg, Germany
2 shared publications- Christos Sachpekidis
Clinical Cooperation Unit Nuclear Medicine, German Cancer Research Center (DKFZ), Im Neuenheimer Feld 280, 69210, Heidelberg, Germany.
2 shared publications
Latest Publications
- Open Access
- 26 August 2026
- Published by Springer Nature in Journal of Imaging Informatics in Medicine
- p. 1-10
Attention-Guided TransMorph for Real-Time Tumor Tracking in Cine-MRI
Magnetic Resonance Imaging (MRI) is a key modality in cancer treatment, providing high soft tissue contrast for the visualization of tumors and internal anatomy. Radiotherapy, which is widely used in treatments, requires precise tumor segmentation to ensure targeting the true tumor and minimizing radiation exposure to healthy tissues. In this regard, real-time automatic tumor tracking from cine-MRI can provide accurate tumor localization supporting adaptive radiotherapy. Conventional image registration techniques exhibit limitations when handling large misalignments and high computational demands, unlike deep learning methods, with high learning capabilities and fast inference times. A real-time tumor tracking approach for 2D cine-MRI using deep learning-based deformable image registration, based on an improved TransMorph architecture, is presented. The approach adheres to a two-step training paradigm: (1) unsupervised pretraining on unlabeled patient image pairs, and (2) supervised fine-tuning with segmentation labels. Attention gates are integrated into skip connections to enhance spatial selectivity regarding the most relevant regions for alignment. A composite loss function is utilized, synthesizing boundary-weighted Dice, adaptive MSE, L1 and smooth diffusion and edge-based regularization. Both overlap and distance-based metrics were computed to assess the model’s segmentation accuracy in the registration of various frames within the patients. The proposed model achieved DSC $$93.42\% \pm 5\%$$ , 95HD $$2.75 \pm 3.24$$ mm, 50HD $$0.88 \pm 0.40$$ mm. Extensive benchmarking demonstrated that the proposed framework consistently achieved superior performance compared to the TransMorph model, its variants, and other existing state-of-the-art image registration approaches. Experimental results indicate that this technique may serve as a critical tool for the advancement of MRI-guided radiotherapy.- 26 August 2026
- Published by Springer Nature in Journal of Imaging Informatics in Medicine
- p. 1-10
Open AccessAttention-Guided TransMorph for Real-Time Tumor Tracking in Cine-MRI
Magnetic Resonance Imaging (MRI) is a key modality in cancer treatment, providing high soft tissue contrast for the visualization of tumors and internal anatomy. Radiotherapy, which is widely used in treatments, requires precise tumor segmentation to ensure targeting the true tumor and minimizing radiation exposure to healthy tissues. In this regard, real-time automatic tumor tracking from cine-MRI can provide accurate tumor localization supporting adaptive radiotherapy. Conventional image registration techniques exhibit limitations when handling large misalignments and high computational demands, unlike deep learning methods, with high learning capabilities and fast inference times. A real-time tumor tracking approach for 2D cine-MRI using deep learning-based deformable image registration, based on an improved TransMorph architecture, is presented. The approach adheres to a two-step training paradigm: (1) unsupervised pretraining on unlabeled patient image pairs, and (2) supervised fine-tuning with segmentation labels. Attention gates are integrated into skip connections to enhance spatial selectivity regarding the most relevant regions for alignment. A composite loss function is utilized, synthesizing boundary-weighted Dice, adaptive MSE, L1 and smooth diffusion and edge-based regularization. Both overlap and distance-based metrics were computed to assess the model’s segmentation accuracy in the registration of various frames within the patients. The proposed model achieved DSC $$93.42\% \pm 5\%$$ , 95HD $$2.75 \pm 3.24$$ mm, 50HD $$0.88 \pm 0.40$$ mm. Extensive benchmarking demonstrated that the proposed framework consistently achieved superior performance compared to the TransMorph model, its variants, and other existing state-of-the-art image registration approaches. Experimental results indicate that this technique may serve as a critical tool for the advancement of MRI-guided radiotherapy. - 15 August 2026
- Published by Springer Nature
- p. 61-69
Boundary-Aware Deformable Registration for MRI-Guided Radiotherapy
- 15 August 2026
- Published by Springer Nature
- p. 61-69
Boundary-Aware Deformable Registration for MRI-Guided Radiotherapy
- 15 August 2026
- Published by Springer Nature
- p. 101-113
A Data-Driven Workflow for Real-Time Hand Tracking in Surgical Video
- 15 August 2026
- Published by Springer Nature
- p. 101-113
A Data-Driven Workflow for Real-Time Hand Tracking in Surgical Video
- 7 October 2025
- Published by Springer Nature in Journal of Imaging Informatics in Medicine
- Vol. 39 (4) , 1-22
U-Swing: An Adaptive U-Net and Swin Fusion for WB-MRI Whole Spine Bone Marrow Segmentation
Whole-body MRI (WB-MRI) is a non-invasive imaging technique offering comprehensive anatomical coverage and high-resolution contrast, ideal for evaluating multi-system diseases without ionizing radiation. Recent advancements in parallel imaging have enhanced its utility in oncology and non-oncology applications. WB-MRI is routinely used for cancer staging, including in multiple myeloma (MM), prostate, and colorectal cancer, as well as in evaluating cancer predisposition syndromes and inflammatory conditions. In MM, WB-MRI is crucial for assessing bone marrow involvement and monitoring treatment response. However, manual analysis of WB-MRI for bone marrow (BM) diseases is time-consuming and prone to data loss, limiting its clinical utility. Tumor load in MM is spatially heterogeneous, requiring detailed BM feature extraction—such as size, volume, intensity, and texture—across the entire bone marrow space. Current guidelines, including Myeloma Response Assessment and Diagnosis System (MY-RADS), offer limited interpretation analysis, and automated methods for comprehensive BM characterisation remain underexplored. These goals rely on automated BM segmentation as a foundational step. This study introduces U-Swing, a hybrid deep learning model combining Swin Transformer (SM) and U-Net Modules (UM) designed for WB-MRI whole spine bone marrow segmentation. U-Swing incorporates dynamic feature fusion of the SMs and UMs via U-Swing Patch Fusion and hierarchical optimization through Stage-Wise U-Swing Adaptation (SUA). The model demonstrated superior performance in WB-MRI bone marrow segmentation using T1-weighted turbo spin-echo (T1W-TSE) sequences, achieving a Dice Similarity (DS) score of 0.928, a Hausdorff Distance (HD95) of 3.919 mm, and an Average Symmetric Surface Distance (ASSD) of 0.281 mm, outperforming model architectures such as U-Net, Swin-UNETR, and UNETR.
- 7 October 2025
- Published by Springer Nature in Journal of Imaging Informatics in Medicine
- Vol. 39 (4) , 1-22
U-Swing: An Adaptive U-Net and Swin Fusion for WB-MRI Whole Spine Bone Marrow Segmentation
Whole-body MRI (WB-MRI) is a non-invasive imaging technique offering comprehensive anatomical coverage and high-resolution contrast, ideal for evaluating multi-system diseases without ionizing radiation. Recent advancements in parallel imaging have enhanced its utility in oncology and non-oncology applications. WB-MRI is routinely used for cancer staging, including in multiple myeloma (MM), prostate, and colorectal cancer, as well as in evaluating cancer predisposition syndromes and inflammatory conditions. In MM, WB-MRI is crucial for assessing bone marrow involvement and monitoring treatment response. However, manual analysis of WB-MRI for bone marrow (BM) diseases is time-consuming and prone to data loss, limiting its clinical utility. Tumor load in MM is spatially heterogeneous, requiring detailed BM feature extraction—such as size, volume, intensity, and texture—across the entire bone marrow space. Current guidelines, including Myeloma Response Assessment and Diagnosis System (MY-RADS), offer limited interpretation analysis, and automated methods for comprehensive BM characterisation remain underexplored. These goals rely on automated BM segmentation as a foundational step. This study introduces U-Swing, a hybrid deep learning model combining Swin Transformer (SM) and U-Net Modules (UM) designed for WB-MRI whole spine bone marrow segmentation. U-Swing incorporates dynamic feature fusion of the SMs and UMs via U-Swing Patch Fusion and hierarchical optimization through Stage-Wise U-Swing Adaptation (SUA). The model demonstrated superior performance in WB-MRI bone marrow segmentation using T1-weighted turbo spin-echo (T1W-TSE) sequences, achieving a Dice Similarity (DS) score of 0.928, a Hausdorff Distance (HD95) of 3.919 mm, and an Average Symmetric Surface Distance (ASSD) of 0.281 mm, outperforming model architectures such as U-Net, Swin-UNETR, and UNETR.
- 5 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Medical Imaging
- Vol. PP (02780062) , 1
Representation learning in PET scans enhanced by semantic and 3D position specific characteristics
Representation learning methods that discover task and/or data-specific characteristics are very popular for a variety of applications. However, their application to 3D medical images is restricted by the computational cost and their inherent subtle differences in intensities and appearance. In this paper, a novel representation learning scheme for extracting representations capable of distinguishing high-uptake regions from 3D 18F-Fluorodeoxyglucose positron emission tomography (FDG-PET) images is proposed. In particular, we propose a novel position-enhanced learning scheme effectively incorporating semantic and position-based features through our proposed Position Encoding Block (PEB) to produce highly informative representations. Such representations incorporate both semantic and position-aware features from high-dimensional medical data, leading to general representations with better performance on clinical tasks. To evaluate our method, we conducted experiments on the challenging task of classifying high-uptake regions as either non-tumor or tumor lesions in Metastatic Melanoma (MM). MM is a type of cancer characterized by its rapid spread to various body sites, which leads to low survival rates. Extensive experiments on an in-house and a public dataset of wholebody FDG-PET images indicated an increase of 10.50% in sensitivity and 4.89% in F1-score against the baseline representation learning scheme while also outperforming state-of-the-art methods for classifying MM regions of interest. The source code will be available at https://github. com/theoVag/Representation-Learning-Sem-Pos.
- 5 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Medical Imaging
- Vol. PP (02780062) , 1
Representation learning in PET scans enhanced by semantic and 3D position specific characteristics
Representation learning methods that discover task and/or data-specific characteristics are very popular for a variety of applications. However, their application to 3D medical images is restricted by the computational cost and their inherent subtle differences in intensities and appearance. In this paper, a novel representation learning scheme for extracting representations capable of distinguishing high-uptake regions from 3D 18F-Fluorodeoxyglucose positron emission tomography (FDG-PET) images is proposed. In particular, we propose a novel position-enhanced learning scheme effectively incorporating semantic and position-based features through our proposed Position Encoding Block (PEB) to produce highly informative representations. Such representations incorporate both semantic and position-aware features from high-dimensional medical data, leading to general representations with better performance on clinical tasks. To evaluate our method, we conducted experiments on the challenging task of classifying high-uptake regions as either non-tumor or tumor lesions in Metastatic Melanoma (MM). MM is a type of cancer characterized by its rapid spread to various body sites, which leads to low survival rates. Extensive experiments on an in-house and a public dataset of wholebody FDG-PET images indicated an increase of 10.50% in sensitivity and 4.89% in F1-score against the baseline representation learning scheme while also outperforming state-of-the-art methods for classifying MM regions of interest. The source code will be available at https://github. com/theoVag/Representation-Learning-Sem-Pos.
- Open Access
- 13 December 2024
- Published by MDPI AG in Electronics
- Vol. 13 (24) , 4919
Automatic Segmentation in 3D CT Images: A Comparative Study of Deep Learning Architectures for the Automatic Segmentation of the Abdominal Aorta
Abdominal aortic aneurysm (AAA) is a complex vascular condition associated with high mortality rates. Accurate abdominal aorta segmentation is essential in medical imaging, facilitating diagnosis and treatment for a range of cardiovascular diseases. In this regard, deep learning-based automated segmentation has shown significant promise in the precise delineation of the aorta. However, comparisons across different models remain limited, with most studies performing algorithmic training and testing on the same dataset. Furthermore, due to the variability in AAA presentation, using healthy controls for deep learning AAA segmentation poses a significant challenge. This study provides a detailed comparative analysis of four deep learning architectures—UNet, SegResNet, UNet Transformers (UNETR), and Shifted-Windows UNet Transformers (SwinUNETR)—for full abdominal aorta segmentation. The models were evaluated both qualitatively and quantitatively using private and public 3D (Computed Tomography) CT datasets. Moreover, they were successful in attaining high performance in delineating AAA aorta, while being trained on healthy aortic imaging data. Our findings indicate that the UNet architecture achieved the highest segmentation accuracy among the models tested.
- 13 December 2024
- Published by MDPI AG in Electronics
- Vol. 13 (24) , 4919
Open AccessAutomatic Segmentation in 3D CT Images: A Comparative Study of Deep Learning Architectures for the Automatic Segmentation of the Abdominal Aorta
Abdominal aortic aneurysm (AAA) is a complex vascular condition associated with high mortality rates. Accurate abdominal aorta segmentation is essential in medical imaging, facilitating diagnosis and treatment for a range of cardiovascular diseases. In this regard, deep learning-based automated segmentation has shown significant promise in the precise delineation of the aorta. However, comparisons across different models remain limited, with most studies performing algorithmic training and testing on the same dataset. Furthermore, due to the variability in AAA presentation, using healthy controls for deep learning AAA segmentation poses a significant challenge. This study provides a detailed comparative analysis of four deep learning architectures—UNet, SegResNet, UNet Transformers (UNETR), and Shifted-Windows UNet Transformers (SwinUNETR)—for full abdominal aorta segmentation. The models were evaluated both qualitatively and quantitatively using private and public 3D (Computed Tomography) CT datasets. Moreover, they were successful in attaining high performance in delineating AAA aorta, while being trained on healthy aortic imaging data. Our findings indicate that the UNet architecture achieved the highest segmentation accuracy among the models tested.
- Open Access
- 5 September 2024
- Published by MDPI AG in Electronics
- Vol. 13 (17) , 3526
A Multidimensional Framework Incorporating 2D U-Net and 3D Attention U-Net for the Segmentation of Organs from 3D Fluorodeoxyglucose-Positron Emission Tomography Images
Accurate analysis of Fluorodeoxyglucose (FDG)-Positron Emission Tomography (PET) images is crucial for the diagnosis, treatment assessment, and monitoring of patients suffering from various cancer types. FDG-PET images provide valuable insights by revealing regions where FDG, a glucose analog, accumulates within the body. While regions of high FDG uptake include suspicious tumor lesions, FDG also accumulates in non-tumor-specific regions and organs. Identifying these regions is crucial for excluding them from certain measurements, or calculating useful parameters, for example, the mean standardized uptake value (SUV) to assess the metabolic activity of the liver. Manual organ delineation from FDG-PET by clinicians demands significant effort and time, which is often not feasible in real clinical workflows with high patient loads. For this reason, this study focuses on automatically identifying key organs with high FDG uptake, namely the brain, left cardiac ventricle, kidneys, liver, and bladder. To this end, an ensemble approach is adopted, where a three-dimensional Attention U-Net (3D AU-Net) is employed for robust three-dimensional analysis, while a two-dimensional U-Net (2D U-Net) is utilized for analysis in the coronal plane. The 3D AU-Net demonstrates highly detailed organ segmentations, but also includes many false positive regions. In contrast, 2D U-Net achieves higher reliability with minimal false positive regions, but lacks the 3D details. Experiments conducted on a subset of the public AutoPET dataset with 60 PET scans demonstrate that the proposed ensemble model achieves high accuracy in segmenting the required organs, surpassing current state-of-the-art techniques, and supporting the potential utilization of the proposed methodology in accelerating and enhancing the clinical workflow of cancer patients.
- 5 September 2024
- Published by MDPI AG in Electronics
- Vol. 13 (17) , 3526
Open AccessA Multidimensional Framework Incorporating 2D U-Net and 3D Attention U-Net for the Segmentation of Organs from 3D Fluorodeoxyglucose-Positron Emission Tomography Images
Accurate analysis of Fluorodeoxyglucose (FDG)-Positron Emission Tomography (PET) images is crucial for the diagnosis, treatment assessment, and monitoring of patients suffering from various cancer types. FDG-PET images provide valuable insights by revealing regions where FDG, a glucose analog, accumulates within the body. While regions of high FDG uptake include suspicious tumor lesions, FDG also accumulates in non-tumor-specific regions and organs. Identifying these regions is crucial for excluding them from certain measurements, or calculating useful parameters, for example, the mean standardized uptake value (SUV) to assess the metabolic activity of the liver. Manual organ delineation from FDG-PET by clinicians demands significant effort and time, which is often not feasible in real clinical workflows with high patient loads. For this reason, this study focuses on automatically identifying key organs with high FDG uptake, namely the brain, left cardiac ventricle, kidneys, liver, and bladder. To this end, an ensemble approach is adopted, where a three-dimensional Attention U-Net (3D AU-Net) is employed for robust three-dimensional analysis, while a two-dimensional U-Net (2D U-Net) is utilized for analysis in the coronal plane. The 3D AU-Net demonstrates highly detailed organ segmentations, but also includes many false positive regions. In contrast, 2D U-Net achieves higher reliability with minimal false positive regions, but lacks the 3D details. Experiments conducted on a subset of the public AutoPET dataset with 60 PET scans demonstrate that the proposed ensemble model achieves high accuracy in segmenting the required organs, surpassing current state-of-the-art techniques, and supporting the potential utilization of the proposed methodology in accelerating and enhancing the clinical workflow of cancer patients.
- Open Access
- 24 February 2024
- Published by MDPI AG in Bioengineering
- Vol. 11 (3) , 214
Towards Automation in Radiotherapy Planning: A Deep Learning Approach for the Delineation of Parotid Glands in Head and Neck Cancer
The delineation of parotid glands in head and neck (HN) carcinoma is critical to assess radiotherapy (RT) planning. Segmentation processes ensure precise target position and treatment precision, facilitate monitoring of anatomical changes, enable plan adaptation, and enhance overall patient safety. In this context, artificial intelligence (AI) and deep learning (DL) have proven exceedingly effective in precisely outlining tumor tissues and, by extension, the organs at risk. This paper introduces a DL framework using the AttentionUNet neural network for automatic parotid gland segmentation in HN cancer. Extensive evaluation of the model is performed in two public and one private dataset, while segmentation accuracy is compared with other state-of-the-art DL segmentation schemas. To assess replanning necessity during treatment, an additional registration method is implemented on the segmentation output, aligning images of different modalities (Computed Tomography (CT) and Cone Beam CT (CBCT)). AttentionUNet outperforms similar DL methods (Dice Similarity Coefficient: 82.65% ± 1.03, Hausdorff Distance: 6.24 mm ± 2.47), confirming its effectiveness. Moreover, the subsequent registration procedure displays increased similarity, providing insights into the effects of RT procedures for treatment planning adaptations. The implementation of the proposed methods indicates the effectiveness of DL not only for automatic delineation of the anatomical structures, but also for the provision of information for adaptive RT support.
- 24 February 2024
- Published by MDPI AG in Bioengineering
- Vol. 11 (3) , 214
Open AccessTowards Automation in Radiotherapy Planning: A Deep Learning Approach for the Delineation of Parotid Glands in Head and Neck Cancer
The delineation of parotid glands in head and neck (HN) carcinoma is critical to assess radiotherapy (RT) planning. Segmentation processes ensure precise target position and treatment precision, facilitate monitoring of anatomical changes, enable plan adaptation, and enhance overall patient safety. In this context, artificial intelligence (AI) and deep learning (DL) have proven exceedingly effective in precisely outlining tumor tissues and, by extension, the organs at risk. This paper introduces a DL framework using the AttentionUNet neural network for automatic parotid gland segmentation in HN cancer. Extensive evaluation of the model is performed in two public and one private dataset, while segmentation accuracy is compared with other state-of-the-art DL segmentation schemas. To assess replanning necessity during treatment, an additional registration method is implemented on the segmentation output, aligning images of different modalities (Computed Tomography (CT) and Cone Beam CT (CBCT)). AttentionUNet outperforms similar DL methods (Dice Similarity Coefficient: 82.65% ± 1.03, Hausdorff Distance: 6.24 mm ± 2.47), confirming its effectiveness. Moreover, the subsequent registration procedure displays increased similarity, providing insights into the effects of RT procedures for treatment planning adaptations. The implementation of the proposed methods indicates the effectiveness of DL not only for automatic delineation of the anatomical structures, but also for the provision of information for adaptive RT support.
- 10 February 2024
- Published by Springer Nature
- p. 80-94
Deep Learning-Based Segmentation and Mesh Reconstruction of the Aortic Vessel Tree from CTA Images
- 10 February 2024
- Published by Springer Nature
- p. 80-94
Deep Learning-Based Segmentation and Mesh Reconstruction of the Aortic Vessel Tree from CTA Images
- 2 May 2023
- Published by SPIE-Intl Soc Optical Eng in Journal of Medical Imaging
- Vol. 10 (03) , 034002-24
Deep learning-based registration of two-dimensional dental images with edge specific loss
PurposeImage registration is a very common procedure in dental applications for aligning images. Registration between pairs of images taken from different angles can improve diagnosis. Our study presents an edge-enhanced unsupervised deep learning (DL)-based deformable registration framework for aligning two-dimensional (2D) pairs of dental x-ray images.ApproachThe proposed neural network is based on the combination of a U-Net like structure, which produces a displacement field, combined with spatial transformer networks, which produce the transformed image. The proposed structure is trained end-to-end by minimizing a weighted loss function consisting of three parts corresponding to image similarity, edge similarity, and registration restrictions. In this regard, the proposed edge specific loss enhances the unsupervised training of the registration framework without the need of supervision through anatomical structures.ResultsThe proposed framework was applied to two datasets, a set of 104 x-ray images of mandibles, arranged in 2600 pairs for training and testing and a set of 17 pairs of pre- and post-operative reconstructed panoramic images. The proposed model outperformed both conventional registration methods and DL-based techniques for both qualitative and quantitative assessment, in most of the compared metrics concerning intensity similarity and edge distances.ConclusionsThe proposed framework achieved accurate and fast deformable alignment of pairs of 2D dental radiographic images. The edge-based module of the loss function enhances the unsupervised learning by directing the network toward deformations that take into consideration the edges of the depicted objects (teeth, bone, and tissue), which are crucial in diagnosis.
- 2 May 2023
- Published by SPIE-Intl Soc Optical Eng in Journal of Medical Imaging
- Vol. 10 (03) , 034002-24
Deep learning-based registration of two-dimensional dental images with edge specific loss
PurposeImage registration is a very common procedure in dental applications for aligning images. Registration between pairs of images taken from different angles can improve diagnosis. Our study presents an edge-enhanced unsupervised deep learning (DL)-based deformable registration framework for aligning two-dimensional (2D) pairs of dental x-ray images.ApproachThe proposed neural network is based on the combination of a U-Net like structure, which produces a displacement field, combined with spatial transformer networks, which produce the transformed image. The proposed structure is trained end-to-end by minimizing a weighted loss function consisting of three parts corresponding to image similarity, edge similarity, and registration restrictions. In this regard, the proposed edge specific loss enhances the unsupervised training of the registration framework without the need of supervision through anatomical structures.ResultsThe proposed framework was applied to two datasets, a set of 104 x-ray images of mandibles, arranged in 2600 pairs for training and testing and a set of 17 pairs of pre- and post-operative reconstructed panoramic images. The proposed model outperformed both conventional registration methods and DL-based techniques for both qualitative and quantitative assessment, in most of the compared metrics concerning intensity similarity and edge distances.ConclusionsThe proposed framework achieved accurate and fast deformable alignment of pairs of 2D dental radiographic images. The edge-based module of the loss function enhances the unsupervised learning by directing the network toward deformations that take into consideration the edges of the depicted objects (teeth, bone, and tissue), which are crucial in diagnosis.