Xinghua Xu

Dr. Xinghua Xu

orcid

Research Focus & Metrics

Academic Career Length

9Years (2018-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
24 February 2021 in Brain and Behavior
Verified
Journal Article
Open access
PDF file
XXXinghua Xu
Xinghua Xu
JZJiashu Zhang
Jiashu Zhang
KYKai Yang
Kai Yang
QWQun Wang
Qun Wang
XCXiaolei Chen
Xiaolei Chen
BXBainan Xu
Bainan Xu
Objectives: Spontaneous intracerebral hemorrhage remains a major cause of death and disability throughout the world. We tried to establish accurate long‐term outcome prediction models for hypertensive intracerebral hemorrhage (HICH) using CT radiomics and machine learning.Methods: In a retrospective study of 270 patients with HICH between June 2013 and June 2018, CT images and patients' 6‐month outcome based on the modified Rankin Scale were collected. Hematomas on CT images were selected as volumes of interests (VOIs), and 1,029 radiomics features of the VOIs were extracted. Based on correlations with patients' outcome, radiomics features underwent dimensionality reduction analyses. Then, the support vector machine (SVM), k‐nearest neighbor (KNN), logistic regression (LR), decision tree (DT), random forest (RF), and XGBoost algorithms were applied with the screened features to establish prognostic prediction models of HICH. Accuracies of all models were compared.Results: Eighteen radiomics features were screened as prognosis‐associated radiomics signature of HICH based on the variance threshold, SelectKBest, and least absolute shrinkage and selection operator (LASSO) regression models. Patients were randomly allocated into training (n = 215) and validation (n = 55) sets. Accuracies of all 6 machine learning algorithms in the validation set exceeded 80%. The sensitivity, specificity, and accuracy in the validation set were 93.3%, 92.5%, and 92.7% for the RF model and 92.3%, 88.1%, and 89.1% for the XGBoost model, respectively, which were the best two among all models.Conclusions: Taking advantage of radiomics and machine learning, we established accurate prognostic prediction models of HICH. The RF model and XGBoost model returned the best accuracies.

13 June 2024 in BMC Medicine
Verified
Journal Article
Open access
PDF file
XXXinghua Xu
Xinghua Xu
HZHuaping Zhang
Huaping Zhang
JZJiashu Zhang
Jiashu Zhang
MLMing Luo
Ming Luo
QWQun Wang
Qun Wang
TTthe MISICH study team
the MISICH study team
Background: Intracerebral hemorrhage (ICH) is a common stroke type with high morbidity and mortality. There are mainly three surgical methods for treating ICH. Unfortunately, thus far, no specific surgical method has been proven to be the most effective. We carried out this study to investigate whether minimally invasive surgeries with endoscopic surgery or stereotactic aspiration (frameless navigated aspiration) will improve functional outcomes in patients with supratentorial ICH compared with small-bone flap craniotomy. Methods: In this parallel-group multicenter randomized controlled trial conducted at 16 centers, patients with supratentorial hypertensive ICH were randomized to receive endoscopic surgery, stereotactic aspiration, or craniotomy at a 1:1:1 ratio from July 2016 to June 2022. The follow-up duration was 6 months. Patients were randomized to receive endoscopic evacuation, stereotactic aspiration, or small-bone flap craniotomy. The primary outcome was favorable functional outcome, defined as the proportion of patients who achieved a modified Rankin scale (mRS) score of 0–2 at the 6-month follow-up. Results: A total of 733 patients were randomly allocated to three groups: 243 to the endoscopy group, 247 to the aspiration group, and 243 to the craniotomy group. Finally, 721 patients (239 in the endoscopy group, 246 in the aspiration group, and 236 in the craniotomy group) received treatment and were included in the intention-to-treat analysis. Primary efficacy analysis revealed that 73 of 219 (33.3%) in the endoscopy group, 72 of 220 (32.7%) in the aspiration group, and 47 of 212 (22.2%) in the craniotomy group achieved favorable functional outcome at the 6-month follow-up (P = .017). We got similar results in subgroup analysis of deep hemorrhages, while in lobar hemorrhages the prognostic outcome was similar among three groups. Old age, deep hematoma location, large hematoma volume, low preoperative GCS score, craniotomy, and intracranial infection were associated with greater odds of unfavorable outcomes. The mean hospitalization expenses were ¥92,420 in the endoscopy group, ¥77,351 in the aspiration group, and ¥100,947 in the craniotomy group (P = .000). Conclusions: Compared with small bone flap craniotomy, endoscopic surgery and stereotactic aspiration improved the long-term outcome of hypertensive ICH, especially deep hemorrhages. Trial Registration: ClinicalTrials.gov Identifier: NCT02811614.

24 March 2018 in Journal of Neuro-Oncology
Verified
Journal Article
Closed access
PDF file
QWQun Wang
Qun Wang
JZJiashu Zhang
Jiashu Zhang
XXXinghua Xu
Xinghua Xu
XCXiaoLei Chen
XiaoLei Chen
BXBainan Xu
Bainan Xu

7 November 2023 in Bioengineering
Verified
Journal Article
Open access
PDF file
Ziyu QiZiyu Qi
Ziyu Qi
Miriam H. A. BoppMiriam H. A. Bopp
Miriam H. A. Bopp
Christopher NimskyChristopher Nimsky
Christopher Nimsky
XCXiaolei Chen
Xiaolei Chen
XXXinghua Xu
Xinghua Xu
JZJiashu Zhang
Jiashu Zhang
Mixed Reality Navigation (MRN) is pivotal in augmented reality-assisted intelligent neurosurgical interventions. However, existing MRN registration methods face challenges in concurrently achieving low user dependency, high accuracy, and clinical applicability. This study proposes and evaluates a novel registration method based on a laser crosshair simulator, evaluating its feasibility and accuracy. A novel registration method employing a laser crosshair simulator was introduced, designed to replicate the scanner frame’s position on the patient. The system autonomously calculates the transformation, mapping coordinates from the tracking space to the reference image space. A mathematical model and workflow for registration were designed, and a Universal Windows Platform (UWP) application was developed on HoloLens-2. Finally, a head phantom was used to measure the system’s target registration error (TRE). The proposed method was successfully implemented, obviating the need for user interactions with virtual objects during the registration process. Regarding accuracy, the average deviation was 3.7 ± 1.7 mm. This method shows encouraging results in efficiency and intuitiveness and marks a valuable advancement in low-cost, easy-to-use MRN systems. The potential for enhancing accuracy and adaptability in intervention procedures positions this approach as promising for improving surgical outcomes.

1 December 2020
Verified
Conference Paper
Closed access
YLYing-Jie Liang
Ying-Jie Liang
XCXiao-Peng Cui
Xiao-Peng Cui
XXXinghua Xu
Xinghua Xu
FJFeng Jiang
Feng Jiang
Object detection has received great attention in recent years. Traditional machine learning object detection pays more attention to extract interpretable features. Deep learning uses CNN for self-learning features, which greatly improves the accuracy. This article summarized and analyzed the existing deep learning object detection method from one-stage and two-stage object detection methods. The roadmap of deep learning target detection is comprehensively analyzed. We have compared the detection performance of typical deep learning target detection algorithms. Finally, aiming at the problems and development trend of machine vision, we make prediction and prospect for the application and development of object detection in the future.

Verified
Journal Article
Open access
PDF file
XXXinghua Xu
Xinghua Xu
JZJiashu Zhang
Jiashu Zhang
ZGZhichao Gan
Zhichao Gan
SZShiyu Zhang
Shiyu Zhang
HZHaoyang Zheng
Haoyang Zheng
QWQun Wang
Qun Wang
Intracerebral hemorrhage (ICH) remains one of the most severe forms of stroke and is associated with high mortality, poor functional outcomes, and substantial healthcare burden worldwide. Despite advances in neurocritical care and minimally invasive surgical techniques, the management of ICH remains challenging because of disease heterogeneity, rapid neurological deterioration, and the lack of effective individualized treatment strategies. In recent years, artificial intelligence (AI) has emerged as a promising tool for improving the diagnosis, prognostic evaluation, and precision management of ICH. A systematic literature search was performed in PubMed, Web of Science, and Embase to identify studies on AI applications in ICH, with predefined inclusion criteria focusing on imaging analysis, prognostic prediction, clinical decision support, and minimally invasive surgery. This review summarizes the major clinical applications, limitations, and future directions of AI in ICH, including multimodal foundation models, intelligent surgical assistance, and personalized precision care. Recent studies have demonstrated that AI-based models can significantly improve the accuracy of hematoma segmentation, hematoma expansion prediction, and functional outcome prognostication compared with conventional approaches. Despite encouraging progress, several important barriers continue to limit clinical translation, including data heterogeneity, limited external validation, insufficient interpretability, ethical and regulatory concerns, and challenges in workflow integration. Overall, AI has the potential to transform ICH management from conventional experience-based practice to data-driven, personalized, and precision neurosurgical care.