Cheng-Ting Shih
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
Cheng-Ting Shih has a total of 109 co-authors. Here are the top co-authors by number of shared publications:
Jay Wu
Department of Biomedical Imaging and Radiological Sciences, National Yang-Ming University, Taipei, Taiwan
10 shared publicationsShin-Lei Peng
6 shared publicationsShu-Jun Chang
Atomic Energy Council, Institute of Nuclear Energy Research, Health Physics Division, No. 1000, Wenhua Road, Jiaan Village, Longtan Township, Taoyuan, Taiwan
6 shared publications- Tung-Hsin Wu
Department of Biomedical Imaging and Radiological Sciences, National Yang Ming Chiao Tung University, Taipei, Taiwan
4 shared publications Yi-Wen Chen
4 shared publications- Yuan-Jen Chang
Department of Aerospace and Systems Engineering, Feng Chia University, Taichung City 407102, Taiwan
3 shared publications
Latest Publications
- 1 September 2026
- Published by Elsevier in Radiation Physics and Chemistry
- Vol. 251
Phantomless patient-specific stoichiometric calibration for deriving stopping power ratios from computed tomography images
- 1 September 2026
- Published by Elsevier in Radiation Physics and Chemistry
- Vol. 251
Phantomless patient-specific stoichiometric calibration for deriving stopping power ratios from computed tomography images
- 19 September 2025
- Published by Elsevier in Computers in Biology and Medicine
- Vol. 197 (Pt B) , 111111
Enhancing the reliability of Alzheimer's disease prediction in MRI images
- 19 September 2025
- Published by Elsevier in Computers in Biology and Medicine
- Vol. 197 (Pt B) , 111111
Enhancing the reliability of Alzheimer's disease prediction in MRI images
- 25 August 2025
- Published by Elsevier in Computers in Biology and Medicine
- Vol. 197 (Pt A) , 110987
Synthesizing whole-body CT images from Dixon MR images using tissue composition prior knowledge and a parametric physical model
- 25 August 2025
- Published by Elsevier in Computers in Biology and Medicine
- Vol. 197 (Pt A) , 110987
Synthesizing whole-body CT images from Dixon MR images using tissue composition prior knowledge and a parametric physical model
- Open Access
- 8 April 2025
- Published by Springer Nature in BMC Cardiovascular Disorders
- Vol. 25 (1) , 1-12
Predicting complications and morbidities in PAD patients through lower extremity compositions with dual-energy CT and material decomposition: a 2-year follow-up observational study
Peripheral artery disease (PAD) is associated with various morbidities. This study aims to investigate the correlation between different lower extremity compositions and development of morbidities in PAD patients. Between January 2018 and December 2020, 108 subjects diagnosed of PAD were enrolled (mean age of 64.1 ± 13.5 years) and utilized dual-energy computed tomography (DECT) with material decomposition to measure the vessel volume, muscle volume, fat volume, and cortical-bone volume in lower extremity respectively. The association between each leg composition and developing complications or morbidities in PAD patients was analyzed over a two-year follow-up. Fontaine stage 3 and 4 had lower muscle mass compared to stages 1 and 2. More severe vascular stenosis was associated with lower muscle, fat, and cortical-bone volume. Patients with severe Fontaine stages (3 and 4) and lower-leg vascular stenosis had a higher risk of developing infection or inflammation (OR 45.5, 95% CI: 13.5–166.7, and OR 11.7, 95% CI: 2.8–50, P < 0.05) and amputation (OR 18.2, 95% CI: 2.2–142.8, and OR 10.7, 95% CI: 1.11–100, P < 0.05). Lower thigh cortical-bone volume was associated with an increased risk of falls resulting in fractures (OR: 1.39, 95% CI: 1.13–2.19, P < 0.01). Thigh cortical-bone volume below 64.5 cm3 was identified as the cut-off value to predict fall-related fractures, with a sensitivity of 100% and specificity of 92%. This study demonstrates the potential of DECT with material decomposition to assess lower extremity composition and its relevance in predicting complications and morbidities in PAD patients. Severe vascular stenosis may contribute to muscle wasting and subsequent complications, while lower thigh cortical-bone mass may serve as a predictor of fall-related fractures.
- 8 April 2025
- Published by Springer Nature in BMC Cardiovascular Disorders
- Vol. 25 (1) , 1-12
Open AccessPredicting complications and morbidities in PAD patients through lower extremity compositions with dual-energy CT and material decomposition: a 2-year follow-up observational study
Peripheral artery disease (PAD) is associated with various morbidities. This study aims to investigate the correlation between different lower extremity compositions and development of morbidities in PAD patients. Between January 2018 and December 2020, 108 subjects diagnosed of PAD were enrolled (mean age of 64.1 ± 13.5 years) and utilized dual-energy computed tomography (DECT) with material decomposition to measure the vessel volume, muscle volume, fat volume, and cortical-bone volume in lower extremity respectively. The association between each leg composition and developing complications or morbidities in PAD patients was analyzed over a two-year follow-up. Fontaine stage 3 and 4 had lower muscle mass compared to stages 1 and 2. More severe vascular stenosis was associated with lower muscle, fat, and cortical-bone volume. Patients with severe Fontaine stages (3 and 4) and lower-leg vascular stenosis had a higher risk of developing infection or inflammation (OR 45.5, 95% CI: 13.5–166.7, and OR 11.7, 95% CI: 2.8–50, P < 0.05) and amputation (OR 18.2, 95% CI: 2.2–142.8, and OR 10.7, 95% CI: 1.11–100, P < 0.05). Lower thigh cortical-bone volume was associated with an increased risk of falls resulting in fractures (OR: 1.39, 95% CI: 1.13–2.19, P < 0.01). Thigh cortical-bone volume below 64.5 cm3 was identified as the cut-off value to predict fall-related fractures, with a sensitivity of 100% and specificity of 92%. This study demonstrates the potential of DECT with material decomposition to assess lower extremity composition and its relevance in predicting complications and morbidities in PAD patients. Severe vascular stenosis may contribute to muscle wasting and subsequent complications, while lower thigh cortical-bone mass may serve as a predictor of fall-related fractures.
- 4 March 2024
- Published by Springer Nature in Physical and Engineering Sciences in Medicine
- Vol. 47 (2) , 755-767
Pattern classification of interstitial lung diseases from computed tomography images using a ResNet-based network with a split-transform-merge strategy and split attention
In patients with interstitial lung disease (ILD), accurate pattern assessment from their computed tomography (CT) images could help track lung abnormalities and evaluate treatment efficacy. Based on excellent image classification performance, convolutional neural networks (CNNs) have been massively investigated for classifying and labeling pathological patterns in the CT images of ILD patients. However, previous studies rarely considered the three-dimensional (3D) structure of the pathological patterns of ILD and used two-dimensional network input. In addition, ResNet-based networks such as SE-ResNet and ResNeXt with high classification performance have not been used for pattern classification of ILD. This study proposed a SE-ResNeXt-SA-18 for classifying pathological patterns of ILD. The SE-ResNeXt-SA-18 integrated the multipath design of the ResNeXt and the feature weighting of the squeeze-and-excitation network with split attention. The classification performance of the SE-ResNeXt-SA-18 was compared with the ResNet-18 and SE-ResNeXt-18. The influence of the input patch size on classification performance was also evaluated. Results show that the classification accuracy was increased with the increase of the patch size. With a 32 × 32 × 16 input, the SE-ResNeXt-SA-18 presented the highest performance with average accuracy, sensitivity, and specificity of 0.991, 0.979, and 0.994. High-weight regions in the class activation maps of the SE-ResNeXt-SA-18 also matched the specific pattern features. In comparison, the performance of the SE-ResNeXt-SA-18 is superior to the previously reported CNNs in classifying the ILD patterns. We concluded that the SE-ResNeXt-SA-18 could help track or monitor the progress of ILD through accuracy pattern classification.
- 4 March 2024
- Published by Springer Nature in Physical and Engineering Sciences in Medicine
- Vol. 47 (2) , 755-767
Pattern classification of interstitial lung diseases from computed tomography images using a ResNet-based network with a split-transform-merge strategy and split attention
In patients with interstitial lung disease (ILD), accurate pattern assessment from their computed tomography (CT) images could help track lung abnormalities and evaluate treatment efficacy. Based on excellent image classification performance, convolutional neural networks (CNNs) have been massively investigated for classifying and labeling pathological patterns in the CT images of ILD patients. However, previous studies rarely considered the three-dimensional (3D) structure of the pathological patterns of ILD and used two-dimensional network input. In addition, ResNet-based networks such as SE-ResNet and ResNeXt with high classification performance have not been used for pattern classification of ILD. This study proposed a SE-ResNeXt-SA-18 for classifying pathological patterns of ILD. The SE-ResNeXt-SA-18 integrated the multipath design of the ResNeXt and the feature weighting of the squeeze-and-excitation network with split attention. The classification performance of the SE-ResNeXt-SA-18 was compared with the ResNet-18 and SE-ResNeXt-18. The influence of the input patch size on classification performance was also evaluated. Results show that the classification accuracy was increased with the increase of the patch size. With a 32 × 32 × 16 input, the SE-ResNeXt-SA-18 presented the highest performance with average accuracy, sensitivity, and specificity of 0.991, 0.979, and 0.994. High-weight regions in the class activation maps of the SE-ResNeXt-SA-18 also matched the specific pattern features. In comparison, the performance of the SE-ResNeXt-SA-18 is superior to the previously reported CNNs in classifying the ILD patterns. We concluded that the SE-ResNeXt-SA-18 could help track or monitor the progress of ILD through accuracy pattern classification.
- Open Access
- 24 April 2023
- Published by Springer Nature in Insights into Imaging
- Vol. 14 (1) , 1-10
Machine learning combined with radiomics and deep learning features extracted from CT images: a novel AI model to distinguish benign from malignant ovarian tumors
Background: To develop an artificial intelligence (AI) model with radiomics and deep learning (DL) features extracted from CT images to distinguish benign from malignant ovarian tumors. Methods: We enrolled 149 patients with pathologically confirmed ovarian tumors. A total of 185 tumors were included and divided into training and testing sets in a 7:3 ratio. All tumors were manually segmented from preoperative contrast-enhanced CT images. CT image features were extracted using radiomics and DL. Five models with different combinations of feature sets were built. Benign and malignant tumors were classified using machine learning (ML) classifiers. The model performance was compared with five radiologists on the testing set. Results: Among the five models, the best performing model is the ensemble model with a combination of radiomics, DL, and clinical feature sets. The model achieved an accuracy of 82%, specificity of 89% and sensitivity of 68%. Compared with junior radiologists averaged results, the model had a higher accuracy (82% vs 66%) and specificity (89% vs 65%) with comparable sensitivity (68% vs 67%). With the assistance of the model, the junior radiologists achieved a higher average accuracy (81% vs 66%), specificity (80% vs 65%), and sensitivity (82% vs 67%), approaching to the performance of senior radiologists. Conclusions: We developed a CT-based AI model that can differentiate benign and malignant ovarian tumors with high accuracy and specificity. This model significantly improved the performance of less-experienced radiologists in ovarian tumor assessment, and may potentially guide gynecologists to provide better therapeutic strategies for these patients.
- 24 April 2023
- Published by Springer Nature in Insights into Imaging
- Vol. 14 (1) , 1-10
Open AccessMachine learning combined with radiomics and deep learning features extracted from CT images: a novel AI model to distinguish benign from malignant ovarian tumors
Background: To develop an artificial intelligence (AI) model with radiomics and deep learning (DL) features extracted from CT images to distinguish benign from malignant ovarian tumors. Methods: We enrolled 149 patients with pathologically confirmed ovarian tumors. A total of 185 tumors were included and divided into training and testing sets in a 7:3 ratio. All tumors were manually segmented from preoperative contrast-enhanced CT images. CT image features were extracted using radiomics and DL. Five models with different combinations of feature sets were built. Benign and malignant tumors were classified using machine learning (ML) classifiers. The model performance was compared with five radiologists on the testing set. Results: Among the five models, the best performing model is the ensemble model with a combination of radiomics, DL, and clinical feature sets. The model achieved an accuracy of 82%, specificity of 89% and sensitivity of 68%. Compared with junior radiologists averaged results, the model had a higher accuracy (82% vs 66%) and specificity (89% vs 65%) with comparable sensitivity (68% vs 67%). With the assistance of the model, the junior radiologists achieved a higher average accuracy (81% vs 66%), specificity (80% vs 65%), and sensitivity (82% vs 67%), approaching to the performance of senior radiologists. Conclusions: We developed a CT-based AI model that can differentiate benign and malignant ovarian tumors with high accuracy and specificity. This model significantly improved the performance of less-experienced radiologists in ovarian tumor assessment, and may potentially guide gynecologists to provide better therapeutic strategies for these patients.
- Open Access
- 10 April 2023
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 18 (4) , e0282737
Evaluating the contact anatomy and contact bone volume of spinal screws using a novel drilled surface image
Intraoperative navigation systems have been widely applied in spinal fusion surgery to improve the implantation accuracy of spinal screws using orthogonal tomographic and surface-rendering imaging. However, these images contain limited anatomical information and no information on bone volume contact by the implanted screw, which has been proven to affect the stability of implanted screws. This study proposed a novel drilled surface imaging technique that displays anatomical integration properties to calculate the contact bone volume (CBV) of the screws implanted along an implantation trajectory. A cylinder was used to represent the area traversed by the screws, which was manually rotated and translated to a predetermined implantation trajectory according to a vertebra model obtained using computed tomography (CT) image volumes. The drilled surface image was reconstructed by interpolating the CT numbers at the predefined sampling points on the cylinder surface. The anatomical integration property and CBV of the screw implanted along the transpedicular trajectory (TT) and cortical bone trajectory (CBT) were evaluated and compared. The drilled surface image fully revealed the contact anatomical structure of the screw under the trajectories, improving the understanding of the anatomical integration of the screw and surrounding tissues. On average, the CBV of the CBT was 30% greater than that of the TT. The proposed drilled surface image may be applied in preoperative planning and integrated into intraoperative navigation systems to evaluate the anatomical integration and degree of bone contact of the screw implanted along a trajectory.
- 10 April 2023
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 18 (4) , e0282737
Open AccessEvaluating the contact anatomy and contact bone volume of spinal screws using a novel drilled surface image
Intraoperative navigation systems have been widely applied in spinal fusion surgery to improve the implantation accuracy of spinal screws using orthogonal tomographic and surface-rendering imaging. However, these images contain limited anatomical information and no information on bone volume contact by the implanted screw, which has been proven to affect the stability of implanted screws. This study proposed a novel drilled surface imaging technique that displays anatomical integration properties to calculate the contact bone volume (CBV) of the screws implanted along an implantation trajectory. A cylinder was used to represent the area traversed by the screws, which was manually rotated and translated to a predetermined implantation trajectory according to a vertebra model obtained using computed tomography (CT) image volumes. The drilled surface image was reconstructed by interpolating the CT numbers at the predefined sampling points on the cylinder surface. The anatomical integration property and CBV of the screw implanted along the transpedicular trajectory (TT) and cortical bone trajectory (CBT) were evaluated and compared. The drilled surface image fully revealed the contact anatomical structure of the screw under the trajectories, improving the understanding of the anatomical integration of the screw and surrounding tissues. On average, the CBV of the CBT was 30% greater than that of the TT. The proposed drilled surface image may be applied in preoperative planning and integrated into intraoperative navigation systems to evaluate the anatomical integration and degree of bone contact of the screw implanted along a trajectory.
- 1 November 2021
- Published by American Scientific Publishers in Journal of Medical Imaging and Health Informatics
- Vol. 11 (11) , 2701-2708
Solving the Prosthesis Modeling for Skull Repair Through Differential Evolution Algorithm
Cranial defects can result in compromised physical protection for the brain and a how risky the brain infection is. Cranioplasty is commonly performed by doing the bone graft onto the deficient area or areas and/or using the metal to support them for restoring the cranial cavity integrity and maintain the physiological intracranial pressure stability. Nowadays, the suitable shape of skull prosthesis can be created and operated precisely and efficiently during cranioplasty process, because the technological development of additive manufacturing or 3D printing. Additive manufacturing has great potential in regard to addressing irregular cranial defects because it can be used to create customized shapes rapidly. However, an unsuitable cranial prosthesis that made from synthetic polymer or a metal implantation will cause a serious infections, and required additional surgery. This paper proposes a geometric model of skull defects by using the superellipse and Differential Evolution (DE). The defects of skill bones in each tomography slice can be modeled by superellipse. The DE optimizes the parameters of superellipse to emulate and compensate the suitable curvature. In a rapid 2D image process and 3D cranial model construction system, the clinical surgeons’ ability is determining, processing, and implanting a customized prosthesis for patients just in a short time in surgery and with maximum surgical quality, especially in emergency cases.
- 1 November 2021
- Published by American Scientific Publishers in Journal of Medical Imaging and Health Informatics
- Vol. 11 (11) , 2701-2708
Solving the Prosthesis Modeling for Skull Repair Through Differential Evolution Algorithm
Cranial defects can result in compromised physical protection for the brain and a how risky the brain infection is. Cranioplasty is commonly performed by doing the bone graft onto the deficient area or areas and/or using the metal to support them for restoring the cranial cavity integrity and maintain the physiological intracranial pressure stability. Nowadays, the suitable shape of skull prosthesis can be created and operated precisely and efficiently during cranioplasty process, because the technological development of additive manufacturing or 3D printing. Additive manufacturing has great potential in regard to addressing irregular cranial defects because it can be used to create customized shapes rapidly. However, an unsuitable cranial prosthesis that made from synthetic polymer or a metal implantation will cause a serious infections, and required additional surgery. This paper proposes a geometric model of skull defects by using the superellipse and Differential Evolution (DE). The defects of skill bones in each tomography slice can be modeled by superellipse. The DE optimizes the parameters of superellipse to emulate and compensate the suitable curvature. In a rapid 2D image process and 3D cranial model construction system, the clinical surgeons’ ability is determining, processing, and implanting a customized prosthesis for patients just in a short time in surgery and with maximum surgical quality, especially in emergency cases.
- Top Cited Papers
- 4 May 2021
- Published by Taylor & Francis in Virtual and Physical Prototyping
- Vol. 16 (3) , 372-386
Perspectives of using machine learning in laser powder bed fusion for metal additive manufacturing
The adoption of laser powder bed fusion (L-PBF) for metals by the industry has been limited despite the significant progress made in the development of the process chain. One of the key obstacles is the inconsistency of the parts obtained from L-PBF. Due to its complexity, there are many potential fluctuations that can occur within the process chain which can lead to quality inconsistency in L-PBF parts. Machine learning (ML) has the possibility to overcome this obstacle by utilising datasets obtained at various stages of the L-PBF process chain. In this perspective article, the integration of ML into the different stages of L-PBF process chain, which potentially lead to better quality control, is explored. Prior to L-PBF, ML can be used for part designs and file preparation. Then, ML algorithms can be applied in the process parameter optimisation and in situ monitoring. Finally, ML can also be integrated into the post-processing.
- 4 May 2021
- Published by Taylor & Francis in Virtual and Physical Prototyping
- Vol. 16 (3) , 372-386
Top Cited PapersPerspectives of using machine learning in laser powder bed fusion for metal additive manufacturing
The adoption of laser powder bed fusion (L-PBF) for metals by the industry has been limited despite the significant progress made in the development of the process chain. One of the key obstacles is the inconsistency of the parts obtained from L-PBF. Due to its complexity, there are many potential fluctuations that can occur within the process chain which can lead to quality inconsistency in L-PBF parts. Machine learning (ML) has the possibility to overcome this obstacle by utilising datasets obtained at various stages of the L-PBF process chain. In this perspective article, the integration of ML into the different stages of L-PBF process chain, which potentially lead to better quality control, is explored. Prior to L-PBF, ML can be used for part designs and file preparation. Then, ML algorithms can be applied in the process parameter optimisation and in situ monitoring. Finally, ML can also be integrated into the post-processing.