Kai Xu
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
Kai Xu has a total of 26 co-authors. Here are the top co-authors by number of shared publications:
Qian Xu
Department of Radiology, The Affiliated Hospital of Xuzhou Medical University, Xuzhou City, Jiangsu Province 221002, China
4 shared publications- Qian Xu
Department of Radiology, the Affiliated Hospital of Xuzhou Medical University, Xuzhou 221002, China
2 shared publications Cailuan Lu
Department of Radiology
1 shared publicationChun Yang
Affiliated Hospital of Xuzhou Medical College
1 shared publicationChunjuan Jiang
Wuxi No.2 People’s Hospital
1 shared publicationHaitao Ge
Department of Medical Imaging, Xuzhou Medical University
1 shared publication
Latest Publications
- 17 October 2022
- Published by Elsevier in European Journal of Radiology
- Vol. 157, 110560
A multitask classification framework based on vision transformer for predicting molecular expressions of glioma
- 17 October 2022
- Published by Elsevier in European Journal of Radiology
- Vol. 157, 110560
A multitask classification framework based on vision transformer for predicting molecular expressions of glioma
- Open Access
- 8 August 2022
- Published by MDPI AG in Journal of Clinical Medicine
- Vol. 11 (15) , 4625
Swin Transformer Improves the IDH Mutation Status Prediction of Gliomas Free of MRI-Based Tumor Segmentation
Background: Deep learning (DL) could predict isocitrate dehydrogenase (IDH) mutation status from MRIs. Yet, previous work focused on CNNs with refined tumor segmentation. To bridge the gap, this study aimed to evaluate the feasibility of developing a Transformer-based network to predict the IDH mutation status free of refined tumor segmentation. Methods: A total of 493 glioma patients were recruited from two independent institutions for model development (TCIA; N = 259) and external test (AHXZ; N = 234). IDH mutation status was predicted directly from T2 images with a Swin Transformer and conventional ResNet. Furthermore, to investigate the necessity of refined tumor segmentation, seven strategies for the model input image were explored: (i) whole tumor slice; (ii–iii) tumor mask and/or not edema; (iv–vii) tumor bounding box of 0.8, 1.0, 1.2, 1.5 times. Performance comparison was made among the networks of different architectures along with different image input strategies, using area under the curve (AUC) and accuracy (ACC). Finally, to further boost the performance, a hybrid model was built by incorporating the images with clinical features. Results: With the seven proposed input strategies, seven Swin Transformer models and seven ResNet models were built, respectively. Based on the seven Swin Transformer models, an averaged AUC of 0.965 (internal test) and 0.842 (external test) were achieved, outperforming 0.922 and 0.805 resulting from the seven ResNet models, respectively. When a bounding box of 1.0 times was used, Swin Transformer (AUC = 0.868, ACC = 80.7%), achieved the best results against the one that used tumor segmentation (Tumor + Edema, AUC = 0.862, ACC = 78.5%). The hybrid model that integrated age and location features into images yielded improved performance (AUC = 0.878, Accuracy = 82.0%) over the model that used images only. Conclusions: Swin Transformer outperforms the CNN-based ResNet in IDH prediction. Using bounding box input images benefits the DL networks in IDH prediction and makes the IDH prediction free of refined glioma segmentation feasible.
- 8 August 2022
- Published by MDPI AG in Journal of Clinical Medicine
- Vol. 11 (15) , 4625
Open AccessSwin Transformer Improves the IDH Mutation Status Prediction of Gliomas Free of MRI-Based Tumor Segmentation
Background: Deep learning (DL) could predict isocitrate dehydrogenase (IDH) mutation status from MRIs. Yet, previous work focused on CNNs with refined tumor segmentation. To bridge the gap, this study aimed to evaluate the feasibility of developing a Transformer-based network to predict the IDH mutation status free of refined tumor segmentation. Methods: A total of 493 glioma patients were recruited from two independent institutions for model development (TCIA; N = 259) and external test (AHXZ; N = 234). IDH mutation status was predicted directly from T2 images with a Swin Transformer and conventional ResNet. Furthermore, to investigate the necessity of refined tumor segmentation, seven strategies for the model input image were explored: (i) whole tumor slice; (ii–iii) tumor mask and/or not edema; (iv–vii) tumor bounding box of 0.8, 1.0, 1.2, 1.5 times. Performance comparison was made among the networks of different architectures along with different image input strategies, using area under the curve (AUC) and accuracy (ACC). Finally, to further boost the performance, a hybrid model was built by incorporating the images with clinical features. Results: With the seven proposed input strategies, seven Swin Transformer models and seven ResNet models were built, respectively. Based on the seven Swin Transformer models, an averaged AUC of 0.965 (internal test) and 0.842 (external test) were achieved, outperforming 0.922 and 0.805 resulting from the seven ResNet models, respectively. When a bounding box of 1.0 times was used, Swin Transformer (AUC = 0.868, ACC = 80.7%), achieved the best results against the one that used tumor segmentation (Tumor + Edema, AUC = 0.862, ACC = 78.5%). The hybrid model that integrated age and location features into images yielded improved performance (AUC = 0.878, Accuracy = 82.0%) over the model that used images only. Conclusions: Swin Transformer outperforms the CNN-based ResNet in IDH prediction. Using bounding box input images benefits the DL networks in IDH prediction and makes the IDH prediction free of refined glioma segmentation feasible.
- Open Access
- 21 April 2020
- Published by Springer Nature in Research Square
Radiological changes of serial CT examinations in patients recovering from coronavirus disease 2019 (COVID-19)
Purpose: To find the pulmonary CT imaging characteristics in patients recovering from coronavirus disease 2019 (COVID-19).Method: Twenty patients with confirmed COVID-19 were enrolled. We analyzed the changes of four pulmonary CT imaging manifestations (ground glass opacity, consolidation, crazy paving sign and cord/band sign) in patients during hospitalization. The disease course was divided into four stages: early stage (0-4 days), progressive stage (5-8 days), peak stage (9-13 days) and absorption stage (≥14 days).Results: There were 12 male and 8 female with an average age of 45±16 years. In the first three stages, GGO was the most common sign on CT imaging. Then, the proportion of GGO decreased in the absorption stage compared with the first three stages (PConclusions: CT imaging showed different characteristics during the four stages. The proportion of cord/band sign significantly increased in the third stage, which might be an indicator of COVID-19 improvement.
- 21 April 2020
- Published by Springer Nature in Research Square
Open AccessRadiological changes of serial CT examinations in patients recovering from coronavirus disease 2019 (COVID-19)
Purpose: To find the pulmonary CT imaging characteristics in patients recovering from coronavirus disease 2019 (COVID-19).Method: Twenty patients with confirmed COVID-19 were enrolled. We analyzed the changes of four pulmonary CT imaging manifestations (ground glass opacity, consolidation, crazy paving sign and cord/band sign) in patients during hospitalization. The disease course was divided into four stages: early stage (0-4 days), progressive stage (5-8 days), peak stage (9-13 days) and absorption stage (≥14 days).Results: There were 12 male and 8 female with an average age of 45±16 years. In the first three stages, GGO was the most common sign on CT imaging. Then, the proportion of GGO decreased in the absorption stage compared with the first three stages (PConclusions: CT imaging showed different characteristics during the four stages. The proportion of cord/band sign significantly increased in the third stage, which might be an indicator of COVID-19 improvement.
- Open Access
- 1 December 2017
- Published by Wolters Kluwer Health in Medicine
- Vol. 96 (50) , e9332
Tumor recurrence versus treatment effects in glioma
Gliomas constitute over 90% of primary brain tumors. Accurate identification of glioma recurrence and treatment effects is important, as it can help determine whether to continue with standard adjuvant chemotherapy or to switch to a second-line therapy for recurrence. Our purpose is to compare three dimensional pseudo-continuous arterial spin labeling (3D-pcASL) technique and dynamic susceptibility contrast perfusion magnetic resonance imaging (DSC-MRI) for differentiation tumor recurrence from treatment-related effects in gliomas. Twenty-nine patients with gliomas previously who showed enlarged, contrast-enhancing lesions within the radiation field after surgery and concurrent chemoradiotherapy (CCRT) were assessed with 3D-pcASL and DSC-MRI. These patients were classified into 2 groups, tumor recurrence group (n = 17) and treatment effects group (n = 12), based on pathologic analysis or clinical-radiologic follow-up. The perfusion imaging quality was assessed using a 3-point scale (1 = poor imaging, 2 = moderate imaging, and 3 = good imaging). Comparison for perfusion imaging-quality score between the 2 techniques was performed with Wilcoxon one-sample test. Quantitative analyses were performed between the 2 groups with cerebral blood flow values (ASL-CBF), relative cerebral blood flow values (ASL-rCBF, DSC-rCBF), and relative cerebral blood volume values (DSC-rCBV) using Wilcoxon one-sample test. The intra-class correlation coefficient (ICC) statistics were calculated for testing intrareader variability in regions of interest (ROIs) measurement of all perfusion parameters. The imaging-quality score of 3D-pcASL was higher than that of DSC-MRI (P = .01). The perfusion parameters between tumor recurrence group and treatment effects group had statistically significant differences. There was a significant correlation between ASL-rCBF and DSC-rCBF values (r = 0.803), between ASL-rCBF and DSC-rCBV values (r = 0.763), and between DSC-rCBF and DSC-rCBV (r = 0.907). A receiver operating characteristic (ROC) curve analysis was performed for significant results of perfusion parameters between the 2 groups. Using a cutoff value of 1.110, ASL-rCBF showed the maximum area under the ROC curve (AUC). However, there were no significant differences among different AUCs. The ICC demonstrated excellent agreement for ROIs measurements of ASL-CBF (ICC = 0.9636), dynamic susceptibility contrast- cerebral blood flow (DSC-CBF) (ICC = 0.8508), and dynamic susceptibility contrast-cerebral blood volume (DSC-CBV) (ICC = 0.8543). 3D-pcASL is an alternative perfusion method to DSC-MRI for the differentiation between tumor recurrence and treatment effects in gliomas. 3D-pcASL is noninvasive and shows fewer susceptibility artifacts than DSC-MRI.
- 1 December 2017
- Published by Wolters Kluwer Health in Medicine
- Vol. 96 (50) , e9332
Open AccessTumor recurrence versus treatment effects in glioma
Gliomas constitute over 90% of primary brain tumors. Accurate identification of glioma recurrence and treatment effects is important, as it can help determine whether to continue with standard adjuvant chemotherapy or to switch to a second-line therapy for recurrence. Our purpose is to compare three dimensional pseudo-continuous arterial spin labeling (3D-pcASL) technique and dynamic susceptibility contrast perfusion magnetic resonance imaging (DSC-MRI) for differentiation tumor recurrence from treatment-related effects in gliomas. Twenty-nine patients with gliomas previously who showed enlarged, contrast-enhancing lesions within the radiation field after surgery and concurrent chemoradiotherapy (CCRT) were assessed with 3D-pcASL and DSC-MRI. These patients were classified into 2 groups, tumor recurrence group (n = 17) and treatment effects group (n = 12), based on pathologic analysis or clinical-radiologic follow-up. The perfusion imaging quality was assessed using a 3-point scale (1 = poor imaging, 2 = moderate imaging, and 3 = good imaging). Comparison for perfusion imaging-quality score between the 2 techniques was performed with Wilcoxon one-sample test. Quantitative analyses were performed between the 2 groups with cerebral blood flow values (ASL-CBF), relative cerebral blood flow values (ASL-rCBF, DSC-rCBF), and relative cerebral blood volume values (DSC-rCBV) using Wilcoxon one-sample test. The intra-class correlation coefficient (ICC) statistics were calculated for testing intrareader variability in regions of interest (ROIs) measurement of all perfusion parameters. The imaging-quality score of 3D-pcASL was higher than that of DSC-MRI (P = .01). The perfusion parameters between tumor recurrence group and treatment effects group had statistically significant differences. There was a significant correlation between ASL-rCBF and DSC-rCBF values (r = 0.803), between ASL-rCBF and DSC-rCBV values (r = 0.763), and between DSC-rCBF and DSC-rCBV (r = 0.907). A receiver operating characteristic (ROC) curve analysis was performed for significant results of perfusion parameters between the 2 groups. Using a cutoff value of 1.110, ASL-rCBF showed the maximum area under the ROC curve (AUC). However, there were no significant differences among different AUCs. The ICC demonstrated excellent agreement for ROIs measurements of ASL-CBF (ICC = 0.9636), dynamic susceptibility contrast- cerebral blood flow (DSC-CBF) (ICC = 0.8508), and dynamic susceptibility contrast-cerebral blood volume (DSC-CBV) (ICC = 0.8543). 3D-pcASL is an alternative perfusion method to DSC-MRI for the differentiation between tumor recurrence and treatment effects in gliomas. 3D-pcASL is noninvasive and shows fewer susceptibility artifacts than DSC-MRI.
- Open Access
- 1 August 2016
- Published by Wolters Kluwer Health in Medicine
- Vol. 95 (33) , e4492
Clear cell tumor of the lung
Clear cell tumor of the lung (CCTL) is a rare and mostly benign pulmonary tumor arising from perivascular epithelioid cells. We reported 2 cases of CCTL in a 24-year-old man with chest pain and a 59-year-old man with dyspnea. Their chest CT images revealed a well-defined, homogeneous nodule located in the peripheral region of lung parenchyma. The solid nodule showed intense heterogenous postcontrast enhancement in the arterial phase and a homogenous appearance in the venous phase. The two patients underwent thoracotomy, and pathological examinations, including immunohistochemical studies revealed that both cases were benign clear cell tumors. Due to the fact that only a few cases about CCTL were reported, clinicopathological aspects and radiological characteristics of this type of tumor are not well established. CCTL is a rare and benign primary tumor and majorly occurs in middle-aged and elder people. The finding of a small vessel extending along and partially wrapping up one side of the mass in the CT-enhanced images might be a helpful feature to differentiate CCTL from other lung neoplasms.
- 1 August 2016
- Published by Wolters Kluwer Health in Medicine
- Vol. 95 (33) , e4492
Open AccessClear cell tumor of the lung
Clear cell tumor of the lung (CCTL) is a rare and mostly benign pulmonary tumor arising from perivascular epithelioid cells. We reported 2 cases of CCTL in a 24-year-old man with chest pain and a 59-year-old man with dyspnea. Their chest CT images revealed a well-defined, homogeneous nodule located in the peripheral region of lung parenchyma. The solid nodule showed intense heterogenous postcontrast enhancement in the arterial phase and a homogenous appearance in the venous phase. The two patients underwent thoracotomy, and pathological examinations, including immunohistochemical studies revealed that both cases were benign clear cell tumors. Due to the fact that only a few cases about CCTL were reported, clinicopathological aspects and radiological characteristics of this type of tumor are not well established. CCTL is a rare and benign primary tumor and majorly occurs in middle-aged and elder people. The finding of a small vessel extending along and partially wrapping up one side of the mass in the CT-enhanced images might be a helpful feature to differentiate CCTL from other lung neoplasms.
- 15 May 2014
- Published by Springer Nature in Journal of Molecular Neuroscience
- Vol. 55 (2) , 289-296
The Effects of Fludarabine on Rat Cerebral Ischemia
- 15 May 2014
- Published by Springer Nature in Journal of Molecular Neuroscience
- Vol. 55 (2) , 289-296
The Effects of Fludarabine on Rat Cerebral Ischemia