Beibei Du
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
Beibei Du has a total of 57 co-authors. Here are the top co-authors by number of shared publications:
Alessandro Zalaffi
Neurosurgery Unit of “Dipartimento di Scienze Mediche, Chirurgiche e Neuroscienze” University of Siena, Azienda Ospedaliera Universitaria , Siena , Italy
1 shared publicationAlex Alfieri
Neurosurgery Department, Kantonspital Winterthur, Winterthur, Switzerland
1 shared publicationAlfonso Cerase
Diagnostic and Therapeutic Neuroradiology Department, Siena University Hospital, Siena, Italy
1 shared publicationAndrea Cardia
Neurosurgery Department, Neurocenter of South Switzerland, Lugano, Switzerland
1 shared publicationAndres Rubiano
Neurosciences Institute, El Bosque University, Bogota, Colombia
1 shared publicationAndrew Reisner
Neurosurgery Department, Children’s Healthcare Hospital Atlanta, Atlanta, USA
1 shared publication
Latest Publications
- Open Access
- 10 November 2026
- Published by Springer Nature in Visual Intelligence
- Vol. 4 (1) , 31
Intra- and inter-Mamba for multi-domain hyperspectral image generalization
Hyperspectral image classification, which aims to assign a unique class label to each pixel, has garnered growing attention in recent years. Although many single-source domain adaptation or domain generalization techniques for hyperspectral image classification have been developed, their generalization capability remains limited. This is primarily because single-source domain adaptation methods are typically tailored to a specific target domain, while single-source domain generalization methods have limited access to diverse source domain variations, restricting their applicability to practical scenarios involving multiple datasets or heterogeneous data sources. To address this issue, we propose a novel intra- and inter-Mamba-based multi-domain generalization network called IMDGN for hyperspectral image classification. First, to overcome the limited receptive field of convolutional neural networks and the high computational complexity of Transformers, an intra- and inter-Mamba network was proposed to capture long-range spectral-spatial dependencies in hyperspectral images. Furthermore, to address variations in the spectral response introduced by heterogeneous sensors, we designed a plug-and-play data augmentation (PDA) module. This module generates multi-view hyperspectral data through amplitude scaling and noise injection, thereby improving robustness to illumination variations and noise perturbations. Experiments on four public benchmarks demonstrate the superiority of the proposed method over existing approaches.
- 10 November 2026
- Published by Springer Nature in Visual Intelligence
- Vol. 4 (1) , 31
Open AccessIntra- and inter-Mamba for multi-domain hyperspectral image generalization
Hyperspectral image classification, which aims to assign a unique class label to each pixel, has garnered growing attention in recent years. Although many single-source domain adaptation or domain generalization techniques for hyperspectral image classification have been developed, their generalization capability remains limited. This is primarily because single-source domain adaptation methods are typically tailored to a specific target domain, while single-source domain generalization methods have limited access to diverse source domain variations, restricting their applicability to practical scenarios involving multiple datasets or heterogeneous data sources. To address this issue, we propose a novel intra- and inter-Mamba-based multi-domain generalization network called IMDGN for hyperspectral image classification. First, to overcome the limited receptive field of convolutional neural networks and the high computational complexity of Transformers, an intra- and inter-Mamba network was proposed to capture long-range spectral-spatial dependencies in hyperspectral images. Furthermore, to address variations in the spectral response introduced by heterogeneous sensors, we designed a plug-and-play data augmentation (PDA) module. This module generates multi-view hyperspectral data through amplitude scaling and noise injection, thereby improving robustness to illumination variations and noise perturbations. Experiments on four public benchmarks demonstrate the superiority of the proposed method over existing approaches.
- Top Cited Papers
- 20 September 2025
- Published by Springer Nature in Neurosurgical Review
- Vol. 48 (1) , 1-8
Cranioplasty complications in severe traumatic brain injury: implications of timing of surgery, implant material and incidence of vetriculomegaly versus Post-Traumatic hydrocephalus
Background Despite the increasing number of decompressive craniectomy (DC) in neurotrauma, the optimal timing for elective cranioplasty (CP) is still debated. Little is known about the CP complications related to surgery, implant material, and post-traumatic hydrocephalus. Objectives To explore the correlation between CP timing, implant material, and the incidence of postoperative complications in patients undergoing CP after DC for severe head injuries. Materials and methods A retrospective multicenter study was conducted from January 2010 to December 2021 across 9 European neurosurgical centers. A cohort of 4007 patients who underwent CP following DC for severe head injury was analyzed. Timing was categorized as: ultra-early (< 30 days), early (31–90 days), late (> 90 days). Complications were defined according to Clavien-Dindo classification, requiring revision surgery and/or hospital readmissions. Results Among the 4007 patients, 352 (8.8%) had ultra-early CP, 1627 (40.5%), and 2028 (51.7%) had early and late CP respectively. Cerebrospinal fluid (CSF) derangement was more frequently associated with large defects and the incidence of Sinking Skin Flap Syndrome (SSFS). SSFS was more frequently diagnosed in patients undergoing late surgery whereas hydrocephalus and epilepsy were less frequently encountered in the ultra-early and early groups (p < 0,05). The overall complication rate was 24.6% (985 patients) including internal hydrocephalus (20%), infection (18%), external hydrocephalus (15%), epilepsy (15%), acute extradural (14%) or subdural hematomas (10%), and subdural hygroma (8%). CP stabilized CSF derangement in 80% of cases, which did not progress into overt hydrocephalus, whereas 17% with definite diagnosis of post-traumatic hydrocephalus required a Ventriculo-Peritoneal shunt (VPS). Simultaneous CP and VPS led to infections in all cases, regardless of implant material. Conclusion Surgery timing has a greater impact on CP complications than implant material. CSF derangement represents the single most relevant factor influencing the clinical course of patients undergoing CP.
- 20 September 2025
- Published by Springer Nature in Neurosurgical Review
- Vol. 48 (1) , 1-8
Top Cited PapersCranioplasty complications in severe traumatic brain injury: implications of timing of surgery, implant material and incidence of vetriculomegaly versus Post-Traumatic hydrocephalus
Background Despite the increasing number of decompressive craniectomy (DC) in neurotrauma, the optimal timing for elective cranioplasty (CP) is still debated. Little is known about the CP complications related to surgery, implant material, and post-traumatic hydrocephalus. Objectives To explore the correlation between CP timing, implant material, and the incidence of postoperative complications in patients undergoing CP after DC for severe head injuries. Materials and methods A retrospective multicenter study was conducted from January 2010 to December 2021 across 9 European neurosurgical centers. A cohort of 4007 patients who underwent CP following DC for severe head injury was analyzed. Timing was categorized as: ultra-early (< 30 days), early (31–90 days), late (> 90 days). Complications were defined according to Clavien-Dindo classification, requiring revision surgery and/or hospital readmissions. Results Among the 4007 patients, 352 (8.8%) had ultra-early CP, 1627 (40.5%), and 2028 (51.7%) had early and late CP respectively. Cerebrospinal fluid (CSF) derangement was more frequently associated with large defects and the incidence of Sinking Skin Flap Syndrome (SSFS). SSFS was more frequently diagnosed in patients undergoing late surgery whereas hydrocephalus and epilepsy were less frequently encountered in the ultra-early and early groups (p < 0,05). The overall complication rate was 24.6% (985 patients) including internal hydrocephalus (20%), infection (18%), external hydrocephalus (15%), epilepsy (15%), acute extradural (14%) or subdural hematomas (10%), and subdural hygroma (8%). CP stabilized CSF derangement in 80% of cases, which did not progress into overt hydrocephalus, whereas 17% with definite diagnosis of post-traumatic hydrocephalus required a Ventriculo-Peritoneal shunt (VPS). Simultaneous CP and VPS led to infections in all cases, regardless of implant material. Conclusion Surgery timing has a greater impact on CP complications than implant material. CSF derangement represents the single most relevant factor influencing the clinical course of patients undergoing CP.
- Open Access
- 26 July 2025
- Published by MDPI AG in Algorithms
- Vol. 18 (8) , 468
Document Image Shadow Removal Based on Illumination Correction Method
Due to diverse lighting conditions and photo environments, shadows are almost ubiquitous in images, especially document images captured with mobile devices. Shadows not only seriously affect the visual quality and readability of a document but also significantly hinder image processing. Although shadow removal research has achieved good results in natural scenes, specific studies on document images are lacking. To effectively remove shadows in document images, the dark illumination correction network is proposed, which mainly consists of two modules: shadow detection and illumination correction. First, a simplified shadow-corrected attention block is designed to combine spatial and channel attention, which is used to extract the features, detect the shadow mask, and correct the illumination. Then, the shadow detection block detects shadow intensity and outputs a soft shadow mask to determine the probability of each pixel belonging to shadow. Lastly, the illumination correction block corrects dark illumination with a soft shadow mask and outputs a shadow-free document image. Our experiments on five datasets show that the proposed method achieved state-of-the-art results, proving the effectiveness of illumination correction.
- 26 July 2025
- Published by MDPI AG in Algorithms
- Vol. 18 (8) , 468
Open AccessDocument Image Shadow Removal Based on Illumination Correction Method
Due to diverse lighting conditions and photo environments, shadows are almost ubiquitous in images, especially document images captured with mobile devices. Shadows not only seriously affect the visual quality and readability of a document but also significantly hinder image processing. Although shadow removal research has achieved good results in natural scenes, specific studies on document images are lacking. To effectively remove shadows in document images, the dark illumination correction network is proposed, which mainly consists of two modules: shadow detection and illumination correction. First, a simplified shadow-corrected attention block is designed to combine spatial and channel attention, which is used to extract the features, detect the shadow mask, and correct the illumination. Then, the shadow detection block detects shadow intensity and outputs a soft shadow mask to determine the probability of each pixel belonging to shadow. Lastly, the illumination correction block corrects dark illumination with a soft shadow mask and outputs a shadow-free document image. Our experiments on five datasets show that the proposed method achieved state-of-the-art results, proving the effectiveness of illumination correction.
- 5 July 2025
- Published by Springer Nature in Computational Economics
- Vol. 67 (6) , 4927-4952
Fuzzy Mean-Variance Portfolio Selection Based on Machine Learning
- 5 July 2025
- Published by Springer Nature in Computational Economics
- Vol. 67 (6) , 4927-4952
Fuzzy Mean-Variance Portfolio Selection Based on Machine Learning
- Open Access
- 13 June 2025
- Published by Springer Nature in Supportive Care in Cancer
- Vol. 33 (7) , 1-16
Exploring supportive care needs of lung cancer patients in China and predicting with machine learning models
Purpose This study aims to explore the level of supportive care needs among hospitalized lung cancer patients in China, explore the key influencing factors and use machine learning (ML) to develop predictive models for the level of supportive care needs among hospitalized lung cancer patients. Methods This cross-sectional study collected data on the supportive care needs, demographics, and clinical information of 486 hospitalized lung cancer patients. Univariate and multivariate analyses identified factors associated with these needs. Predictive models were developed using six machine learning methods—logistic regression, linear regression, k-nearest neighbors, support vector machine, random forest, and adaptive boosting—to assess their performance, followed by a visualization of feature importance. The code used for model development and analysis is publicly available at https://github.com/zimengcc/predict_cancer_scn. Results Among the factors influencing the supportive care needs of hospitalized lung cancer patients, age, education level, occupation, tumor stage, and household per capita monthly income have a significant impact on supportive care needs scores. Multiple linear regression analysis revealed that education level and household per capita monthly income were statistically significant predictors of supportive care needs scores. In the predictive tasks, the random forest model performed the best, with a mean absolute error (MAE) of 4.45 for predicting the total supportive care needs score. Furthermore, to predict the dimension with the highest level of supportive care needs, the model achieved an accuracy of 88. 42%, an F1 score of 87. 49%, and an ROC-AUC of 0.9061. Conclusion Our study explored the factors influencing the level of supportive care needs among hospitalized lung cancer patients. While the machine learning models demonstrate promising predictive performance, it is important to note that all results were derived solely through cross-validation. Therefore, potential overfitting and overestimation of model performance should be considered when interpreting these findings. Nevertheless, these models may serve as a foundation for developing tools to support personalized care planning in clinical settings.
- 13 June 2025
- Published by Springer Nature in Supportive Care in Cancer
- Vol. 33 (7) , 1-16
Open AccessExploring supportive care needs of lung cancer patients in China and predicting with machine learning models
Purpose This study aims to explore the level of supportive care needs among hospitalized lung cancer patients in China, explore the key influencing factors and use machine learning (ML) to develop predictive models for the level of supportive care needs among hospitalized lung cancer patients. Methods This cross-sectional study collected data on the supportive care needs, demographics, and clinical information of 486 hospitalized lung cancer patients. Univariate and multivariate analyses identified factors associated with these needs. Predictive models were developed using six machine learning methods—logistic regression, linear regression, k-nearest neighbors, support vector machine, random forest, and adaptive boosting—to assess their performance, followed by a visualization of feature importance. The code used for model development and analysis is publicly available at https://github.com/zimengcc/predict_cancer_scn. Results Among the factors influencing the supportive care needs of hospitalized lung cancer patients, age, education level, occupation, tumor stage, and household per capita monthly income have a significant impact on supportive care needs scores. Multiple linear regression analysis revealed that education level and household per capita monthly income were statistically significant predictors of supportive care needs scores. In the predictive tasks, the random forest model performed the best, with a mean absolute error (MAE) of 4.45 for predicting the total supportive care needs score. Furthermore, to predict the dimension with the highest level of supportive care needs, the model achieved an accuracy of 88. 42%, an F1 score of 87. 49%, and an ROC-AUC of 0.9061. Conclusion Our study explored the factors influencing the level of supportive care needs among hospitalized lung cancer patients. While the machine learning models demonstrate promising predictive performance, it is important to note that all results were derived solely through cross-validation. Therefore, potential overfitting and overestimation of model performance should be considered when interpreting these findings. Nevertheless, these models may serve as a foundation for developing tools to support personalized care planning in clinical settings.
- 7 March 2024
- Published by Springer Nature in Cluster Computing
- Vol. 27 (5) , 6773-6786
A coupled generative graph convolution network by capturing dynamic relationship of regional flow for traffic prediction
- 7 March 2024
- Published by Springer Nature in Cluster Computing
- Vol. 27 (5) , 6773-6786
A coupled generative graph convolution network by capturing dynamic relationship of regional flow for traffic prediction
- 12 October 2023
- Published by Springer Nature in Machine Vision and Applications
- Vol. 34 (6) , 1-10
Simple and effective complementary label learning based on mean square error loss
- 12 October 2023
- Published by Springer Nature in Machine Vision and Applications
- Vol. 34 (6) , 1-10
Simple and effective complementary label learning based on mean square error loss
- 9 August 2023
- Published by Springer Nature in Journal of Applied Mathematics and Computing
- Vol. 69 (5) , 3755-3791
A binary Caputo–Fabrizio fractional reproducing kernel method for the time-fractional Cattaneo equation
- 9 August 2023
- Published by Springer Nature in Journal of Applied Mathematics and Computing
- Vol. 69 (5) , 3755-3791
A binary Caputo–Fabrizio fractional reproducing kernel method for the time-fractional Cattaneo equation
- 5 January 2022
- Published by Springer Nature in Digital Finance
- Vol. 4 (1) , 89-108
Adaptive order flow forecasting with multiplicative error models
- 5 January 2022
- Published by Springer Nature in Digital Finance
- Vol. 4 (1) , 89-108
Adaptive order flow forecasting with multiplicative error models