1
|
Zeng Y, Long C, Zhao W, Liu J. Predicting the Severity of Neurological Impairment Caused by Ischemic Stroke Using Deep Learning Based on Diffusion-Weighted Images. J Clin Med 2022; 11:jcm11144008. [PMID: 35887776 PMCID: PMC9325315 DOI: 10.3390/jcm11144008] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 04/15/2022] [Revised: 06/23/2022] [Accepted: 07/05/2022] [Indexed: 02/01/2023] Open
Abstract
Purpose: To develop a preliminary deep learning model that uses diffusion-weighted imaging (DWI) images to classify the severity of neurological impairment caused by ischemic stroke. Materials and Methods: This retrospective study included 851 ischemic stroke patients (711 patients in the training set and 140 patients in the test set). The patients’ NIHSS scores, which reflect the severity of neurological impairment, were reviewed upon admission and on Day 7 of hospitalization and were classified into two stages (stage 1 for NIHSS < 5 and stage 2 for NIHSS ≥ 5). A 3D-CNN was trained to predict the stage of NIHSS based on different preprocessed DWI images. The performance in predicting the severity of anterior and posterior circulation stroke was also investigated. The AUC, specificity, and sensitivity were calculated to evaluate the performance of the model. Results: Our proposed model obtained better performance in predicting the NIHSS stage on Day 7 of hospitalization than that at admission (best AUC 0.895 vs. 0.846). Model D trained with DWI images (normalized with z-score and resized to 256 × 256 × 64 voxels) achieved the best AUC of 0.846 in predicting the NIHSS stage at admission. Model E rained with DWI images (normalized with maximum−minimum and resized to 128 × 128 × 32 voxels) achieved the best AUC of 0.895 in predicting the NIHSS stage on Day 7 of hospitalization. Our model also showed promising performance in predicting the NIHSS stage on Day 7 of hospitalization for anterior and posterior circulation stroke, with the best AUCs of 0.905 and 0.903, respectively. Conclusions: Our proposed 3D-CNN model can effectively predict the neurological severity of IS using DWI images and performs better in predicting the NIHSS stage on Day 7 of hospitalization. The model also obtained promising performance in subgroup analysis, which can potentially help clinical decision making.
Collapse
Affiliation(s)
- Ying Zeng
- Department of Radiology, The Second Xiangya Hospital, Central South University, Changsha 410011, China;
- Department of Radiology, Xiangtan Central Hospital, Xiangtan 411199, China
| | - Chen Long
- Department of Stroke Unit, Xiangtan Central Hospital, Xiangtan 411199, China;
| | - Wei Zhao
- Department of Radiology, The Second Xiangya Hospital, Central South University, Changsha 410011, China;
- Clinical Research Center for Medical Imaging, Changsha 410011, China
- Correspondence: (W.Z.); (J.L.)
| | - Jun Liu
- Department of Radiology, The Second Xiangya Hospital, Central South University, Changsha 410011, China;
- Clinical Research Center for Medical Imaging, Changsha 410011, China
- Department of Radiology Quality Control Center, Changsha 410011, China
- Correspondence: (W.Z.); (J.L.)
| |
Collapse
|
2
|
Ablefoni M, Ullrich S, Surov A, Hoffmann KT, Meyer HJ. Diagnostic benefit of high b-value computed diffusion-weighted imaging in acute brainstem infarction. J Neuroradiol 2020; 49:47-52. [PMID: 32987036 DOI: 10.1016/j.neurad.2020.09.004] [Citation(s) in RCA: 2] [Impact Index Per Article: 0.5] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 07/09/2020] [Revised: 09/03/2020] [Accepted: 09/13/2020] [Indexed: 11/17/2022]
Abstract
BACKGROUND AND PURPOSE Diffusion-weighted imaging (DWI) is a cornerstone in diagnostic of ischemic stroke. The aim of this study was to investigate the usefulness of high-b-value computed DWI (c-DWI) in comparison to standard DWI in patients with acute brainstem infarction. MATERIALS AND METHODS 56 patients with acute brainstem infarction were retrospectively analysed by two readers. DWI was obtained with the b-values 0, 500 and 1000 s/mm² on either a 1.5 or 3 T magnetic resonance imaging (MRI) scanner. c-DWI was calculated with a monoexponential model with high b-values 2000, 3000, 4000 and 5000 s/mm². All c-DWI series with high-b-values were compared to the standard DWI sequence at b-value of 1000 s/mm² in terms of image artifacts, lesion extent and contrast. RESULTS There was no statistically significant difference between 1.5 and 3 T MRI regarding the measured ischemic lesion size. There were no statistically significant differences between the ischemic lesion sizes on DWI at b-values of 1000 s/mm² and on c-DWI at higher b-values. Overall, the contrast between the lesion and the surrounding normal areas improved with increasing b-value on the isotropic DWIs: maximum at b = 5000, followed by that at b 2000 and b 1000 s/mm², in order. The best relation between artifacts and lesion contrast was identified for b 2000 s/mm². CONCLUSION High b-value DWI derived from c-DWI has a higher visibility for ischemic brainstem lesions compared to standard DWI without additional time cost. The b-2000 image is recommended to use in clinical routine, higher b-value images lead to more imaging artifacts, which might result in misdiagnosis.
Collapse
Affiliation(s)
- Maxime Ablefoni
- Department of Diagnostic and Interventional Radiology, University of Leipzig, Liebigstraße 20, 04103 Leipzig, Germany.
| | - Sebastian Ullrich
- Department of Diagnostic and Interventional Radiology, University of Leipzig, Liebigstraße 20, 04103 Leipzig, Germany
| | - Alexey Surov
- Department of Radiology and Nuclear Medicine, University of Magdeburg, Leipziger Str. 44, 39120 Magdeburg, Germany
| | - Karl-Titus Hoffmann
- Department of Neuroradiology, University of Leipzig, Liebigstraße 20, 04103 Leipzig, Germany
| | - Hans-Jonas Meyer
- Department of Diagnostic and Interventional Radiology, University of Leipzig, Liebigstraße 20, 04103 Leipzig, Germany
| |
Collapse
|
3
|
Takeshige N, Aoki T, Sakata K, Kajiwara S, Negoto T, Nagase S, Tanoue S, Uchiyama Y, Hirohata M, Abe T, Morioka M. Sagittal diffusion-weighted imaging in preventing the false-negative diagnosis of acute brainstem infarction: Confirmation of the benefit by anatomical characterization of false-negative lesions. Surg Neurol Int 2019; 10:180. [PMID: 31637081 PMCID: PMC6778332 DOI: 10.25259/sni_182_2019] [Citation(s) in RCA: 5] [Impact Index Per Article: 1.0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 08/02/2019] [Accepted: 08/08/2019] [Indexed: 11/06/2022] Open
Abstract
Background: In some cases of acute brainstem infarction (BI), standard axial diffusion-weighted imaging (DWI) does not show a lesion, leading to false-negative (FN) diagnoses. It is important to recognize acute BI accurately and promptly to initiate therapy as soon as possible. Methods: Of the 171 patients with acute cerebral infarctions in our institution who were examined, 16 were diagnosed with true-positive BI (TP-BI) and six with FN-BI. We evaluated the effectiveness of sagittal DWI in accurately diagnosing acute BI and sought to find the cause of its effectiveness by the anatomical characterization of FN-BIs. Results: Considering the direction of the brainstem perforating arteries, we supposed that sagittal DWI might more effectively detect BIs than axial DWI. We found that sagittal DWI detected all FN-BIs more clearly than axial DWI. The mean time between the onset of symptoms and initial DWI was significantly longer in the TP group (17.6 ± 5.5 h) than in the FN group (5.0 ± 1.2 h; P < 0.0001). The lesion volumes were much smaller in FN-BIs (259 ± 82 mm3) than in TP-BIs (2779 ± 767 mm3; P = 0.0007). FN-BIs had a significant inverse correlation with the ventrodorsal length of infarcts (FN 3.5 ± 1.1 mm, TP 11.4 ± 3.6 mm; P < 0.0004) and no correlation with other size parameters such as rostrocaudal thickness and lateral width. Conclusion: Anatomical characterization clearly confirmed that the addition of sagittal DWI to the initial axial DWI in suspected cases of BI ensures its accurate diagnosis and improves the patient’s prognosis.
Collapse
Affiliation(s)
- Nobuyuki Takeshige
- Departments of Neurosurgery, Kurume University School of Medicine, Kurume, Fukuoka, Japan
| | - Takachika Aoki
- Departments of Neurosurgery, Kurume University School of Medicine, Kurume, Fukuoka, Japan
| | - Kiyohiko Sakata
- Departments of Neurosurgery, Kurume University School of Medicine, Kurume, Fukuoka, Japan
| | - Soushou Kajiwara
- Departments of Neurosurgery, Kurume University School of Medicine, Kurume, Fukuoka, Japan
| | - Tetsuya Negoto
- Departments of Neurosurgery, Kurume University School of Medicine, Kurume, Fukuoka, Japan
| | - Satoshi Nagase
- Departments of Neurosurgery, Kurume University School of Medicine, Kurume, Fukuoka, Japan
| | - Syuichi Tanoue
- Departments of Radiology, Kurume University School of Medicine, Kurume, Fukuoka, Japan
| | - Yusuke Uchiyama
- Departments of Radiology, Kurume University School of Medicine, Kurume, Fukuoka, Japan
| | - Masaru Hirohata
- Departments of Neurosurgery, Kurume University School of Medicine, Kurume, Fukuoka, Japan
| | - Toshi Abe
- Departments of Radiology, Kurume University School of Medicine, Kurume, Fukuoka, Japan
| | - Motohiro Morioka
- Departments of Neurosurgery, Kurume University School of Medicine, Kurume, Fukuoka, Japan
| |
Collapse
|
4
|
Khaleel NI, Zghair MAG, Hassan QA. Value of Combination of Standard Axial and Thin-Section Coronal Diffusion-weighted Imaging in Diagnosis of Acute Brainstem Infarction. Open Access Maced J Med Sci 2019; 7:2287-2291. [PMID: 31592276 PMCID: PMC6765079 DOI: 10.3889/oamjms.2019.336] [Citation(s) in RCA: 2] [Impact Index Per Article: 0.4] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 03/28/2019] [Revised: 06/04/2019] [Accepted: 06/05/2019] [Indexed: 11/13/2022] Open
Abstract
AIM: To determine the value of the combination of thin-section 3 mm coronal and standard axial DWI and their impact in facilitating the diagnosis of acute brainstem infarction. METHODS: A cross-sectional study conducted from the 1st of April 2017 to the end of February 2018 on 100 consecutive patients (66% were male, and 34% were female) with isolated acute ischemic infarction in the brainstem. The abnormal MRI findings concerning the ischemic lesions were interpreted on standard axial 5 mm and thin-section coronal 3mm DWI. RESULTS: The mean age of the studied group was 69.2 ± 4.3 for male and 72.3 ± 2.5 years. The standard axial DWI can diagnose 20%, 6.7% and 6.7% of the infarctions in midbrain, pons and medulla oblongata respectively, while both axial and thin coronal sections together can diagnose 80% of midbrain infarctions, 93.3% of pons infarctions and 93.3% of medulla oblongata infarctions. Furthermore, the thin section coronal 3 mm section can diagnose very smaller ischemic lesion volume in comparison to the standard axial 5mm section (3.4 ± 0.45 / cm3 versus 4.6 ± 0.23 / cm3, P < 0.001) CONCLUSION: The addition of thin-section coronal DWI can facilitate the detection of brainstem ischemic lesions. We suggest its inclusion in the stroke MRI protocol.
Collapse
Affiliation(s)
- Nashwan I Khaleel
- Department of Radiology, Al-Yarmook Teaching Hospital, Baghdad, Iraq
| | - Muna A G Zghair
- Division of Radiology, Department of Medicine, College of Medicine, Al-Mustansiriyah University, Baghdad, Iraq
| | - Qays A Hassan
- Division of Radiology, Department of Surgery, Al-Kindy College of Medicine, University of Baghdad, Baghdad, Iraq
| |
Collapse
|
5
|
Felfeli P, Wenz H, Al-Zghloul M, Groden C, Förster A. Combination of standard axial and thin-section coronal diffusion-weighted imaging facilitates the diagnosis of brainstem infarction. Brain Behav 2017; 7:e00666. [PMID: 28413710 PMCID: PMC5390842 DOI: 10.1002/brb3.666] [Citation(s) in RCA: 18] [Impact Index Per Article: 2.6] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Submit a Manuscript] [Subscribe] [Scholar Register] [Received: 12/26/2016] [Accepted: 01/25/2017] [Indexed: 11/17/2022] Open
Abstract
BACKGROUND AND PURPOSE Although diffusion-weighted imaging (DWI) is a very sensitive technique for the detection of small ischemic lesions in the human brain, in particular in the brainstem it may fail to demonstrate acute ischemic infarction. In this study, we sought to evaluate the value of additional thin-section coronal DWI for the detection of brainstem infarction. METHODS In 155 consecutive patients (median age 69 [interquartile range, IQR 57-78] years, 95 [61.3%] males) with isolated brainstem infarction, MRI findings were analyzed, with emphasis on ischemic lesions on standard axial (5 mm) and thin-section coronal (3 mm) DWI. RESULTS On DWI, we identified ischemic lesions in the mesencephalon in 12 (7.7%), pons in 115 (74.2%), and medulla oblongata in 31 (20%) patients. In 3 (1.9%) cases-all of these with medulla oblongata infarction-the ischemic lesion was detected only on thin-section coronal DWI. Overall, in 35 (22.6%) patients the ischemic lesion was more easily identified on thin-section coronal DWI in comparison to standard axial DWI. In these, the ischemic lesions were significantly smaller (0.06 [IQR 0.05-0.11] cm3 vs. 0.25 [IQR 0.13-0.47] cm3; p < .001) in comparison to those patients whose ischemic lesion was more easily (6 [3.9%]) or at least similarly well identified (114 [73.5%]) on standard axial DWI. CONCLUSIONS Since thin-section coronal DWI may facilitate the diagnosis of brainstem infarction, we suggest its inclusion in standard stroke MRI protocols.
Collapse
Affiliation(s)
- Philippe Felfeli
- Department of Neuroradiology Universitätsmedizin Mannheim University of Heidelberg Mannheim Germany
| | - Holger Wenz
- Department of Neuroradiology Universitätsmedizin Mannheim University of Heidelberg Mannheim Germany
| | - Mansour Al-Zghloul
- Department of Neuroradiology Universitätsmedizin Mannheim University of Heidelberg Mannheim Germany
| | - Christoph Groden
- Department of Neuroradiology Universitätsmedizin Mannheim University of Heidelberg Mannheim Germany
| | - Alex Förster
- Department of Neuroradiology Universitätsmedizin Mannheim University of Heidelberg Mannheim Germany
| |
Collapse
|
6
|
De Cocker LJL, van Veluw SJ, Fowkes M, Luijten PR, Mali WPTM, Hendrikse J. Very small cerebellar infarcts: integration of recent insights into a functional topographic classification. Cerebrovasc Dis 2013; 36:81-7. [PMID: 24029219 DOI: 10.1159/000353668] [Citation(s) in RCA: 16] [Impact Index Per Article: 1.5] [Reference Citation Analysis] [Abstract] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 03/20/2013] [Accepted: 06/11/2013] [Indexed: 11/19/2022] Open
Abstract
BACKGROUND Very small cerebellar infarcts (diameter <2 cm) are a frequent finding on MRI. With an increasing scientific interest in cerebral microinfarcts, very small infarcts in the cerebellum deserve more of our attention as well. The goal of the present article was to review infarct terminology and mechanisms, as well as to critically appraise the current classification system for very small cerebellar infarcts. METHODS A search strategy was designed to identify all relevant studies on very small cerebellar infarcts in the English language. This search was restricted to papers published up to February 21, 2013. Studies were initially identified from the MEDLINE/PubMed database using the search terms 'small cerebellar infarct', 'lacunar infarct', 'microinfarct', 'end zone infarct', 'border zone infarct', 'watershed infarct', 'territorial infarct', and 'nonterritorial infarct'. Furthermore, a similar search strategy was directed to identify all relevant articles on (descriptive and functional) neuroanatomy and neuroimaging of the cerebellum. RESULTS Very small cerebellar infarcts have been referred to as lacunar infarcts, as junctional, border zone or watershed infarcts, as nonterritorial infarcts, as very small territorial or end zone infarcts, or simply as (very) small cerebellar infarcts. Since the original clinicoradiological study on these small infarcts, the classification into border zones remains in common use. This classification is based upon the assumption that these infarcts occur secondary to low flow in between arterial perfusion territories, where flow is believed to be the lowest. Later studies, however, have suggested occlusion of small (end-) arteries as a prerequisite for the pathogenesis of even small cerebellar infarcts, with low flow merely as a potential contributor. Therefore, it is likely that infarcts may as well occur in a nonborder zone distribution. Moreover, the classification into border zones may be considered unreliable since the location of border zones is highly variable among individuals and is not known in a particular patient. Recently, a functional topographic organization has been found in the cerebellum with evidence for a motor-nonmotor dichotomy between the anterior and posterior lobe. Since the cerebellar lobes can be easily and reliably distinguished with both CT and MRI, we recommend the classification of very small cerebellar infarcts according to topographic location. CONCLUSION There are several fundamental concerns with the current classification of very small cerebellar infarcts according to border zones, which we would like to overcome by recommending a new classification system based on topography. This will allow for a reliable and reproducible way of classifying very small cerebellar infarcts and is expected to improve clinicoradiological correlation.
Collapse
Affiliation(s)
- Laurens J L De Cocker
- Department of Radiology, University Medical Center Utrecht, Utrecht, The Netherlands
| | | | | | | | | | | |
Collapse
|
7
|
Risk Factors for Small Cortical Infarction on Diffusion-Weighted Magnetic Resonance Imaging in Patients With Acute Ischemic Stroke. J Stroke Cerebrovasc Dis 2011; 20:68-74. [DOI: 10.1016/j.jstrokecerebrovasdis.2009.11.003] [Citation(s) in RCA: 4] [Impact Index Per Article: 0.3] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 06/03/2009] [Revised: 10/07/2009] [Accepted: 11/07/2009] [Indexed: 11/20/2022] Open
|
8
|
Chowdhury D, Wardlaw JM, Dennis MS. Are multiple acute small subcortical infarctions caused by embolic mechanisms? J Neurol Neurosurg Psychiatry 2004; 75:1416-20. [PMID: 15377688 PMCID: PMC1738752 DOI: 10.1136/jnnp.2004.038653] [Citation(s) in RCA: 37] [Impact Index Per Article: 1.9] [Reference Citation Analysis] [Abstract] [MESH Headings] [Track Full Text] [Journal Information] [Submit a Manuscript] [Subscribe] [Scholar Register] [Indexed: 11/04/2022]
Abstract
OBJECTIVE To seek evidence of potential embolic sources or other stroke mechanisms in patients who, on chance observation, had several apparently recent small subcortical infarcts on diffusion weighted magnetic resonance imaging (DWI). METHODS Patients presenting with stroke and multiple hyperintense subcortical infarcts visible on DWI were identified prospectively. Detailed clinical and radiological assessments were done independently and blinded to each other. RESULTS Of 10 patients with multiple hyperintense subcortical infarcts on DWI, a definite embolic source was identified in only one. Most patients were hypertensive and smoked. The DWI appearance suggested that the subcortical lesions had occurred within several weeks rather than at exactly the same time. Most patients also had significant white matter hyperintensities and four had microhaemorrhages. CONCLUSIONS Embolic sources were not identified in most patients but they did have systemic vascular risk factors and brain imaging features of "small vessel disease." A more generalised intrinsic process affecting many small cerebral vessels contemporaneously could explain multiple acute small subcortical infarcts. White matter hyperintensities, microhaemorrhages, and multiple small subcortical infarcts may share a common pathophysiological mechanism such as a diffuse cerebral microvascular abnormality which requires further exploration.
Collapse
Affiliation(s)
- D Chowdhury
- Department of Clinical Neurosciences, Western General Hospital, Crewe Road, Edinburgh EH4 2XU, UK.
| | | | | |
Collapse
|
9
|
Abstract
Stroke is a major public-health burden worldwide. Prevention programmes are essential to reduce the incidence of stroke and to prevent the all but inevitable stroke epidemic, which will hit less developed countries particularly hard as their populations age and adopt lifestyles of the more developed countries. Efficient, effective, and rapid diagnosis of stroke and transient ischaemic attack is crucial. The diagnosis of the exact type and cause of stroke, which requires brain imaging as well as traditional clinical skills, is also important when it will influence management. The treatment of acute stroke, the prevention and management of the many complications of stroke, and the prevention of recurrent stroke and other serious vascular events are all improving rapidly. However, stroke management will only be most effective when delivered in the context of an organised, expert, educated, and enthusiastic stroke service that can react quickly to the needs of patients at all stages from onset to recovery.
Collapse
Affiliation(s)
- Charles Warlow
- Division of Clinical Neurosciences, Western General Hospital, EH4 2XU, Edinburgh, UK.
| | | | | | | | | |
Collapse
|
10
|
|
11
|
Affiliation(s)
- J M Wardlaw
- Department of Clinical Neurosciences, Western General Hospital, Edinburgh, UK.
| |
Collapse
|