4 January 2021 in Nano Letters
1 January 2020 in BioMed Research International
Objective. To determine if osteosarcoma (OS) and Ewing sarcoma (EWS) of the pelvis based on MRI can be differentiated using radiomic analysis. Materials and Methods. In this study, 3.0 T magnetic resonance (MR) data of 66 patients (40 males and 26 females, mean age 27.6 ± 13.9 years) with pathologically confirmed OS or EWS of the pelvis (35 with OS and 31 with EWS) taken from April 2013 to December 2017 were retrospectively reviewed. T2‐weighted fat‐saturated (T2‐FS) and contrast‐enhanced T1‐weighted (CET1) images were manually segmented, and imaging features were extracted. Independent‐sample t‐test, Spearman’s test, and the least absolute shrinkage and selection operator (LASSO) method were used to select the most useful features from the original data set. The performance of radiomic analysis was investigated by the area under the receiver operating characteristic (ROC) curve (AUC) analysis. Results. 385 initial features were extracted from T2‐FS and CET1 MR data. Nine features from T2‐FS and 7 features from CET1 were selected by using the LASSO method. The radiomic analysis to differentiate OS and EWS of the pelvis based on T2‐FS and CET1 images using the aforementioned selected features achieved AUC values of 0.881 (95% confidence interval (CI): 0.799–0.963) and 0.765 (95% CI: 0.652–0.878), respectively. Conclusion. Radiomic analysis showed potential in differentiating OS from EWS of the pelvis, in which T2‐FS demonstrated better diagnostic value. To differentiate OS from EWS of the pelvis using our multiparametric MRI‐based radiomic analysis could preoperatively improve diagnostic accuracy and greatly contribute to therapy planning.
7 September 2021 in Frontiers in Aging Neuroscience
Background: Parkinson's disease (PD) is a highly heterogeneous disease, especially in the clinical characteristics and prognosis. The PD is divided into two subgroups: tremor-dominant phenotype and non-tremor-dominant phenotype. Previous studies reported abnormal changes between the two PD phenotypes by using the static functional connectivity analysis. However, the dynamic properties of brain networks between the two PD phenotypes are not yet clear. Therefore, we aimed to uncover the dynamic functional network connectivity (dFNC) between the two PD phenotypes at the subnetwork level, focusing on the temporal properties of dFNC and the variability of network efficiency.Methods: We investigated the resting-state functional MRI (fMRI) data from 29 tremor-dominant PD patients (PDTD), 25 non-tremor-dominant PD patients (PDNTD), and 20 healthy controls (HCs). Sliding window approach, k-means clustering, independent component analysis (ICA), and graph theory analysis were applied to analyze the dFNC. Furthermore, the relationship between alterations in the dynamic properties and clinical features was assessed.Results: The dFNC analyses identified four reoccurring states, one of them showing sparse connections (state I). PDTD patients stayed longer time in state I and showed increased FNC between BG and vSMN in state IV. Both PD phenotypes exhibited higher FNC between dSMN and FPN in state II and state III compared with the controls. PDNTD patients showed decreased FNC between BG and FPN but increased FNC in the bilateral FPN compared with both PDTD patients and controls. In addition, PDNTD patients exhibited greater variability in global network efficiency. Tremor scores were positively correlated with dwell time in state I along with increased FNC between BG and vSMN in state IV.Conclusions: This study explores the dFNC between the PDTD and PDNTD patients, which offers new evidence on the abnormal time-varying brain functional connectivity and their network destruction of the two PD phenotypes, and may help better understand the neural substrates underlying different types of PD.
8 March 2024 in IEEE Journal of Biomedical and Health Informatics
Accurate detection and segmentation of brain tumors is critical for medical diagnosis. However, current supervised learning methods require extensively annotated images and the state-of-the-art generative models used in unsupervised methods often have limitations in covering the whole data distribution. In this paper, we propose a novel framework T wo- S tage G enerative M odel (TSGM) that combines Cycle Generative Adversarial Network (CycleGAN) and V ariance E xploding stochastic differential equation using j oint p robability (VE-JP) to improve brain tumor detection and segmentation. The CycleGAN is trained on unpaired data to generate abnormal images from healthy images as data prior. Then VE-JP is implemented to reconstruct healthy images using synthetic paired abnormal images as a guide, which alters only pathological regions but not regions of healthy. Notably, our method directly learned the joint probability distribution for conditional generation. The residual between input and reconstructed images suggests the abnormalities and a thresholding method is subsequently applied to obtain segmentation results. Furthermore, the multimodal results are weighted with different weights to improve the segmentation accuracy further. We validated our method on three datasets, and compared with other unsupervised methods for anomaly detection and segmentation. The DSC score of 0.8590 in BraTs2020 dataset, 0.6226 in ITCS dataset and 0.7403 in In-house dataset show that our method achieves better segmentation performance and has better generalization.
27 October 2020 in Frontiers in Neurology
Background and Objective: Parkinson disease (PD) with rapid eye movement (REM) sleep behavior disorder (PD-RBD) tend to be a distinct phenotype with more severe clinical characteristics and pathological lesion when compared with PD without RBD (PD-nRBD). However, the pathological mechanism underlying PD-RBD remains unclear. We aim to use the resting-state functional magnetic resonance imaging (rs-fMRI) to explore the mechanism of PD-RBD from the perspective of internal connectivity networks. Materials and Methods: A total of 92 PD patients and 20 age and sex matched normal controls (NC) were included. All participants underwent rs-fMRI scan and clinical assessment. According to the RBD screening questionnaire (RBDSQ), PD patients were divided into two groups: PD with probable RBD (PD-pRBD) and PD without probable RBD (PD-npRBD). The whole brain was divided into 90 regions using automated anatomic labeling atlas. Functional network of each subject was constructed according to the correlation of rs-fMRI blood oxygenation level dependent signals in any two brain regions and network metrics were analyzed using graph theory approaches. Network properties among three groups were compared and correlation analysis was made using distinguishing network metrics and RBDSQ scores. Results: We found both PD-pRBD and PD-npRBD patients existed small-world characteristics. PD-pRBD showed a wider range of nodal property changes in neocortex and limbic system than PD-npRBD patients when compared with NC. Besides, PD-pRBD showed significant enhanced nodal efficiency in the bilateral thalamus and betweenness centrality in the left insula, but, reduced betweenness centrality in the right dorsolateral superior frontal gyrus when compared with PD-npRBD. Moreover, nodal efficiency in the bilateral thalamus were positively correlated with RBDSQ scores. Conclusions: Both NC and PD patients displayed small-world properties and indiscriminate global measure but PD-pRBD showed more extensive changes of nodal properties than PD-npRBD. The increased centrality role in the bilateral thalamus and the left insula, and disruption in the right dorsolateral superior frontal gyrus may play as a key role in underlying pathogenesis of PD-RBD.
1 May 2023 in Quantitative Imaging in Medicine and Surgery
Original Article Deep learning for fully automated segmentation and volumetry of Couinaud liver segments and future liver remnants shown with CT before major hepatectomy: a validation study of a predictive model Tingting Xie, Yongbin Li, Ziying Lin, Xiang Liu, Xiaodong Zhang, Yaofeng Zhang, Dadou Zhang, Guanxun Cheng, Xiaoying Wang