1
|
Zeineldin RA, Karar ME, Elshaer Z, Coburger J, Wirtz CR, Burgert O, Mathis-Ullrich F. Explainable hybrid vision transformers and convolutional network for multimodal glioma segmentation in brain MRI. Sci Rep 2024; 14:3713. [PMID: 38355678 PMCID: PMC10866944 DOI: 10.1038/s41598-024-54186-7] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 06/14/2023] [Accepted: 02/09/2024] [Indexed: 02/16/2024] Open
Abstract
Accurate localization of gliomas, the most common malignant primary brain cancer, and its different sub-region from multimodal magnetic resonance imaging (MRI) volumes are highly important for interventional procedures. Recently, deep learning models have been applied widely to assist automatic lesion segmentation tasks for neurosurgical interventions. However, these models are often complex and represented as "black box" models which limit their applicability in clinical practice. This article introduces new hybrid vision Transformers and convolutional neural networks for accurate and robust glioma segmentation in Brain MRI scans. Our proposed method, TransXAI, provides surgeon-understandable heatmaps to make the neural networks transparent. TransXAI employs a post-hoc explanation technique that provides visual interpretation after the brain tumor localization is made without any network architecture modifications or accuracy tradeoffs. Our experimental findings showed that TransXAI achieves competitive performance in extracting both local and global contexts in addition to generating explainable saliency maps to help understand the prediction of the deep network. Further, visualization maps are obtained to realize the flow of information in the internal layers of the encoder-decoder network and understand the contribution of MRI modalities in the final prediction. The explainability process could provide medical professionals with additional information about the tumor segmentation results and therefore aid in understanding how the deep learning model is capable of processing MRI data successfully. Thus, it enables the physicians' trust in such deep learning systems towards applying them clinically. To facilitate TransXAI model development and results reproducibility, we will share the source code and the pre-trained models after acceptance at https://github.com/razeineldin/TransXAI .
Collapse
Affiliation(s)
- Ramy A Zeineldin
- Department Artificial Intelligence in Biomedical Engineering (AIBE), Friedrich-Alexander-University Erlangen-Nürnberg (FAU), 91052, Erlangen, Germany.
- Research Group Computer Assisted Medicine (CaMed), Reutlingen University, 72762, Reutlingen, Germany.
- Faculty of Electronic Engineering (FEE), Menoufia University, Minuf, 32952, Egypt.
| | - Mohamed E Karar
- Faculty of Electronic Engineering (FEE), Menoufia University, Minuf, 32952, Egypt
| | - Ziad Elshaer
- Department of Neurosurgery, University of Ulm, 89312, Günzburg, Germany
| | - Jan Coburger
- Department of Neurosurgery, University of Ulm, 89312, Günzburg, Germany
| | - Christian R Wirtz
- Department of Neurosurgery, University of Ulm, 89312, Günzburg, Germany
| | - Oliver Burgert
- Research Group Computer Assisted Medicine (CaMed), Reutlingen University, 72762, Reutlingen, Germany
| | - Franziska Mathis-Ullrich
- Department Artificial Intelligence in Biomedical Engineering (AIBE), Friedrich-Alexander-University Erlangen-Nürnberg (FAU), 91052, Erlangen, Germany
| |
Collapse
|
2
|
Knoll A, Durner G, Braun M, Schmitz B, Wirtz CR, König R, Pala A. Combined stent retriever angioplasty and continuous intra-arterial nimodipine infusion as salvage therapy for cerebral vasospasm and delayed cerebral ischemia after subarachnoid hemorrhage: illustrative case. JOURNAL OF NEUROSURGERY. CASE LESSONS 2023; 6:CASE23339. [PMID: 37782962 PMCID: PMC10555600 DOI: 10.3171/case23339] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Subscribe] [Scholar Register] [Received: 06/26/2023] [Accepted: 08/01/2023] [Indexed: 10/04/2023]
Abstract
BACKGROUND Delayed cerebral ischemia (DCI) and cerebral vasospasm (CV) are severe complications of spontaneous subarachnoid hemorrhage (SAH) contributing to an inferior outcome. Rescue therapies include intra-arterial balloon angioplasty and repetitive and finally continuous intra-arterial nimodipine infusion. OBSERVATIONS In the presented case, a young female patient with fulminant refractory DCI and CV, despite induced hypertension and nimodipine application, was treated with three-vessel continuous intra-arterial infusion and additional repetitive angioplasty of the basilar and middle cerebral arteries using a stent retriever, leading to a good clinical outcome. Additional stent retriever dilatation to continuous intra-arterial nimodipine application in three vessel territories may represent a further escalation step in the rescue therapy for severe CV and DCI after SAH. Montreal Cognitive Assessment and SF-36 testing showed satisfactory results 3 months after initial treatment with intra-arterial nimodipine catheters in three vessel territory circulations and additional stent retriever vasodilation of severe CV. LESSONS We report a unique rescue strategy involving implantation of an additional intra-arterial catheter into the vertebral artery and repetitive stent retriever dilatations of the middle cerebral and basilar arteries as an extra therapy for continuous intra-arterial nimodipine vaspospasmolytic therapy in three vessel territories, resulting in a very good clinical outcome.
Collapse
Affiliation(s)
| | | | - Michael Braun
- Neuroradiology, University of Ulm, Günzburg, Germany
| | - Bernd Schmitz
- Neuroradiology, University of Ulm, Günzburg, Germany
| | | | | | | |
Collapse
|
3
|
Han B, Zhang L, Jia W. Contrast-Enhanced Ultrasound in Resection of Spinal Cord Gliomas. World Neurosurg 2023; 171:e83-e92. [PMID: 36427693 DOI: 10.1016/j.wneu.2022.11.087] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 10/12/2022] [Revised: 11/18/2022] [Accepted: 11/19/2022] [Indexed: 11/27/2022]
Abstract
BACKGROUND Intraoperative contrast-enhanced ultrasound (iCEUS) is a relatively new technique for visualizing brain gliomas and can help achieve maximum resection, but its potential in evaluation of spinal cord gliomas has not been well defined. The aim of this study was to describe the iCEUS characterization of and evaluate its role in visualizing intramedullary gliomas. METHODS A retrospective review of patients who underwent intramedullary glioma resection with iCEUS guidance from 2019 to 2021 was conducted. An offline analysis was performed to compare and characterize the perfusion features of each glioma. RESULTS This study included 36 patients who underwent iCEUS for spinal cord gliomas. iCEUS was performed successfully, and all gliomas were clearly identified. The distribution of contrast agent showed different dynamic phases (arterial, peak, and washout) from those observed in brain gliomas, generally appearing slower and less intense in spinal cord gliomas. iCEUS helped highlight intramedullary gliomas, each of which demonstrated specific iCEUS features depending on the grade. Gross total resection was achieved in 20 patients (55.6%), subtotal resection was achieved in 11 patients (30.6%), and partial resection was achieved in 5 patients (13.8%). CONCLUSIONS ICEUS adds valuable information in highlighting spinal cord gliomas in real time. It allows the neurosurgeon to assess the anatomical location of the glioma and delineate the tumor margins. iCEUS could play a potentially important role in guiding spinal cord glioma resection. Further study with more cases is needed to better understand the microbubble distribution dynamics in intramedullary gliomas.
Collapse
Affiliation(s)
- Bo Han
- Department of Neurosurgery, Beijing Tiantan Hospital, Capital Medical University, Beijing, China
| | - Liang Zhang
- National Cancer Center/National Clinical Research Center for Cancer/Hebei Cancer Hospital, Chinese Academy of Medical Sciences, Langfang, China; National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China
| | - Wenqing Jia
- Department of Neurosurgery, Beijing Tiantan Hospital, Capital Medical University, Beijing, China.
| |
Collapse
|
4
|
Yuan T, Gao Z, Wang F, Ren JL, Wang T, Zhong H, Gao G, Quan G. Relative T2-FLAIR signal intensity surrounding residual cavity is associated with survival prognosis in patients with lower-grade gliomas. Front Oncol 2022; 12:960917. [PMID: 36185187 PMCID: PMC9520477 DOI: 10.3389/fonc.2022.960917] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 06/03/2022] [Accepted: 08/23/2022] [Indexed: 11/22/2022] Open
Abstract
Aims To investigate whether the relative signal intensity surrounding the residual cavity on T2-fluid-attenuated inversion recovery (rFLAIR) can improve the survival prediction of lower-grade glioma (LGG) patients. Methods Clinical and pathological data and the follow-up MR imaging of 144 patients with LGG were analyzed. We calculated rFLAIR with Image J software. Logistic analysis was used to explore the significant impact factors on progression-free survival (PFS) and overall survival (OS). Several models were set up to predict the survival prognosis of LGG. Results A higher rFLAIR [1.81 (0.83)] [median (IQR)] of non-enhancing regions surrounding the residual cavity was detected in the progressed group (n=77) than that [1.55 (0.33)] [median (IQR)] of the not-progressed group (n = 67) (P<0.001). Multivariate analysis showed that lower KPS (≤75), and higher rFLAIR (>1.622) were independent predictors for poor PFS (P<0.05), whereas lower KPS (≤75) and thick-linear and nodular enhancement were the independent predictors for poor OS (P<0.05). The cutoff rFLAIR value of 1.622 could be used to predict poor PFS (HR = 0.31, 95%CI 0.20–0.48) (P<0.001) and OS (HR = 0.27, 95%CI 0.14–0.51) (P=0.002). Both the areas under the ROC curve (AUCs) for predicting poor PFS (AUC, 0.771) and OS (AUC, 0.831) with a combined model that contained rFLAIR were higher than those of any other models. Conclusion Higher rFALIR (>1.622) in non-enhancing regions surrounding the residual cavity can be used as a biomarker of the poor survival of LGG. rFLAIR is helpful to improve the survival prediction of posttreatment LGG patients.
Collapse
Affiliation(s)
- Tao Yuan
- Department of Medical Imaging, The Second Hospital of Hebei Medical University, Shijiazhuang, China
| | - Zhen Gao
- Department of Medical Imaging, The Second Hospital of Hebei Medical University, Shijiazhuang, China
| | - Fei Wang
- Department of Medical Imaging, The Second Hospital of Hebei Medical University, Shijiazhuang, China
| | - Jia-Liang Ren
- Department of Pharmaceuticals Diagnostics, General Electric Healthcare China, Beijing, China
| | - Tianda Wang
- Department of Medical Imaging, The Second Hospital of Hebei Medical University, Shijiazhuang, China
| | - Hongbo Zhong
- Department of Radiology, People’s Hospital of Tangshan City, Tangshan, China
| | - Guodong Gao
- Department of Medical Imaging, The Second Hospital of Hebei Medical University, Shijiazhuang, China
| | - Guanmin Quan
- Department of Medical Imaging, The Second Hospital of Hebei Medical University, Shijiazhuang, China
- *Correspondence: Guanmin Quan,
| |
Collapse
|
5
|
Bunyaratavej K, Siwanuwatn R, Tuchinda L, Wangsawatwong P. Impact of Intraoperative Magnetic Resonance Imaging (i-MRI) on Surgeon Decision Making and Clinical Outcomes in Cranial Tumor Surgery. Asian J Neurosurg 2022; 17:218-226. [PMID: 36120606 PMCID: PMC9473858 DOI: 10.1055/s-0042-1751008] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/23/2022] Open
Abstract
Background
Although intraoperative magnetic resonance imaging (iMRI) has an established role in guiding intraoperative extent of resection (EOR) in cranial tumor surgery, the details of how iMRI data are used by the surgeon in the real-time decision-making process is lacking.
Materials and Methods
The authors retrospectively reviewed 40 consecutive patients who underwent cranial tumor resection with the guidance of iMRI. The tumor volumes were measured by volumetric software. Intraoperative and postoperative EOR were calculated and compared. Surgeon preoperative EOR intention, intraoperative EOR assessment, and how iMRI data impacted surgeon decisions were analyzed.
Results
The pathology consisted of 29 gliomas, 8 pituitary tumors, and 3 other tumors. Preoperative surgeon intention called for gross total resection (GTR) in 28 (70%) cases. After resection and before iMRI scanning, GTR was 20 (50.0%) cases based on the surgeon's perception. After iMRI scanning, the results helped identify 19 (47.5%) cases with unexpected results consisting of 5 (12.5%) with unexpected locations of residual tumors and 14 (35%) with unexpected EOR. Additional resection was performed in 24 (60%) cases after iMRI review, including 6 (15%) cases with expected iMRI results. Among 34 cases with postoperative MRI results, iMRI helped improve EOR in 12 (35.3%) cases.
Conclusion
In cranial tumor surgery, the surgeon's preoperative and intraoperative assessment is frequently imprecise. iMRI data serve several purposes, including identifying the presence of residual tumors, providing residual tumor locations, giving spatial relation data of the tumor with nearby eloquent structures, and updating the neuro-navigation system for the final stage of tumor resection.
Collapse
Affiliation(s)
- Krishnapundha Bunyaratavej
- Division of Neurosurgery, Department of Surgery, Faculty of Medicine, Chulalongkorn University and King Chulalongkorn Memorial Hospital, Thai Red Cross Society, Pathumwan, Bangkok, Thailand
| | - Rungsak Siwanuwatn
- Division of Neurosurgery, Department of Surgery, Faculty of Medicine, Chulalongkorn University and King Chulalongkorn Memorial Hospital, Thai Red Cross Society, Pathumwan, Bangkok, Thailand
| | - Lawan Tuchinda
- Department of Anesthesiology, Faculty of Medicine, Chulalongkorn University and King Chulalongkorn Memorial Hospital, Thai Red Cross Society, Pathumwan, Bangkok, Thailand
| | - Piyanat Wangsawatwong
- Division of Neurosurgery, Department of Surgery, Faculty of Medicine, Chulalongkorn University and King Chulalongkorn Memorial Hospital, Thai Red Cross Society, Pathumwan, Bangkok, Thailand
| |
Collapse
|