A deep learning approach versus expert clinician panel in the classification of posterior circulation infarction
Background: Posterior circulation infarction (POCI) is common. Imaging techniques such as non-contrast-CT (NCCT) and diffusion-weighted-magnetic-resonance-imaging commonly fail to detect hyperacute POCI. Studies suggest expert inspection of Computed Tomography Perfusion (CTP) improves diagnosis of P...
Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-01-01
|
| Series: | NeuroImage: Clinical |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2213158225000026 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850035235136208896 |
|---|---|
| author | Leon S. Edwards Milanka Visser Cecilia Cappelen-Smith Dennis Cordato Andrew Bivard Leonid Churilov Christopher Blair James Thomas Angela Dos Santos Longting Lin Chushuang Chen Carlos Garcia-Esperon Kenneth Butcher Tim Kleinig Phillip MC Choi Xin Cheng Qiang Dong Richard I. Aviv Mark W. Parsons |
| author_facet | Leon S. Edwards Milanka Visser Cecilia Cappelen-Smith Dennis Cordato Andrew Bivard Leonid Churilov Christopher Blair James Thomas Angela Dos Santos Longting Lin Chushuang Chen Carlos Garcia-Esperon Kenneth Butcher Tim Kleinig Phillip MC Choi Xin Cheng Qiang Dong Richard I. Aviv Mark W. Parsons |
| author_sort | Leon S. Edwards |
| collection | DOAJ |
| description | Background: Posterior circulation infarction (POCI) is common. Imaging techniques such as non-contrast-CT (NCCT) and diffusion-weighted-magnetic-resonance-imaging commonly fail to detect hyperacute POCI. Studies suggest expert inspection of Computed Tomography Perfusion (CTP) improves diagnosis of POCI. In many settings, there is limited access to specialist expertise. Deep-learning has been successfully applied to automate imaging interpretation. This study aimed to develop and validate a deep-learning approach for the classification of POCI using CTP. Methods: Data were analysed from 3541-patients from the International-stroke-perfusion-registry (INSPIRE). All patients with baseline multimodal-CT and follow-up imaging performed at 24–48 h were identified. A cohort of 541-patients was constructed on a 1:3 POCI-to −reference-ratio for model analysis. A 3D-Dense-Convolutional-Network (DenseNet) was trained to classify patients into POCI or non-POCI using CTP-deconvolved-maps. Six-stroke-experts also independently classified patients based upon stepwise access to multimodal CT (mCT) data. DenseNet results were compared against expert clinician results. Model and clinician performance was evaluated using area-under-the-receiver-operating-curve, sensitivity, specificity, accuracy and precision. Clinician agreement was measured with the Fleiss-Kappa-statistic. Results: Best mean clinician diagnostic accuracy, sensitivity and agreement was demonstrated after review of all mCT data (AUC: 0.81, Sensitivity: 0.65, Fleiss-Kappa-statistic: 0.73). There was a spectrum of individual clinician results with an AUC-range of 0.73–0.86. Best DenseNet performance was recorded with an input combination of NCCT and delay-time maps. The DenseNet model was superior to the best mean clinician performance (AUC: 0.87) and was due to enhanced sensitivity (DenseNET: 0.77, Clinician: 0.65). The degree to which the DenseNet model outperformed each clinician ranged and was clinician specific (AUC improvement 0.01–0.14). Conclusion: Comprehensive review of CTP improves diagnostic performance and agreement amongst clinicians. A DenseNet model was superior to best mean clinician performance. The degree of improvement varied by specific clinician. Development of a clinician-DenseNet approach may improve inter-clinician agreement and diagnostic accuracy. This approach may alleviate limited specialist services in resource constrained settings. |
| format | Article |
| id | doaj-art-9ea8963ceb744fb0b049cf599434ded5 |
| institution | DOAJ |
| issn | 2213-1582 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | Elsevier |
| record_format | Article |
| series | NeuroImage: Clinical |
| spelling | doaj-art-9ea8963ceb744fb0b049cf599434ded52025-08-20T02:57:32ZengElsevierNeuroImage: Clinical2213-15822025-01-014510373210.1016/j.nicl.2025.103732A deep learning approach versus expert clinician panel in the classification of posterior circulation infarctionLeon S. Edwards0Milanka Visser1Cecilia Cappelen-Smith2Dennis Cordato3Andrew Bivard4Leonid Churilov5Christopher Blair6James Thomas7Angela Dos Santos8Longting Lin9Chushuang Chen10Carlos Garcia-Esperon11Kenneth Butcher12Tim Kleinig13Phillip MC Choi14Xin Cheng15Qiang Dong16Richard I. Aviv17Mark W. Parsons18Department of Neurology and Neurophysiology, Liverpool Hospital, Sydney, NSW, Australia; South Western Sydney Clinical School, University of New South Wales, Sydney, NSW, Australia; Ingham Institute for Applied Medical Research, Sydney, NSW, Australia; Corresponding author at: Department of Neurology and Neurophysiology, Liverpool Hospital, Sydney, NSW, Australia.Melbourne Brain Centre at the Royal Melbourne Hospital, University of Melbourne, Parkville, AustraliaDepartment of Neurology and Neurophysiology, Liverpool Hospital, Sydney, NSW, Australia; South Western Sydney Clinical School, University of New South Wales, Sydney, NSW, Australia; Ingham Institute for Applied Medical Research, Sydney, NSW, AustraliaDepartment of Neurology and Neurophysiology, Liverpool Hospital, Sydney, NSW, Australia; South Western Sydney Clinical School, University of New South Wales, Sydney, NSW, Australia; Ingham Institute for Applied Medical Research, Sydney, NSW, AustraliaMelbourne Brain Centre at the Royal Melbourne Hospital, University of Melbourne, Parkville, AustraliaMelbourne Brain Centre at the Royal Melbourne Hospital, University of Melbourne, Parkville, AustraliaSouth Western Sydney Clinical School, University of New South Wales, Sydney, NSW, Australia; Ingham Institute for Applied Medical Research, Sydney, NSW, AustraliaDepartment of Neurology and Neurophysiology, Liverpool Hospital, Sydney, NSW, Australia; South Western Sydney Clinical School, University of New South Wales, Sydney, NSW, Australia; Ingham Institute for Applied Medical Research, Sydney, NSW, AustraliaSouth Western Sydney Clinical School, University of New South Wales, Sydney, NSW, AustraliaDepartment of Neurology and Neurophysiology, Liverpool Hospital, Sydney, NSW, Australia; South Western Sydney Clinical School, University of New South Wales, Sydney, NSW, AustraliaMelbourne Brain Centre at the Royal Melbourne Hospital, University of Melbourne, Parkville, AustraliaDepartment of Neurology, John Hunter Hospital, Newcastle, NSW, Australia; Hunter Medical Research Institute and University of Newcastle, Newcastle, NSW, AustraliaPrince of Wales Clinical School, University of New South Wales, Sydney, NSW, AustraliaDepartment of Neurology, Royal Adelaide Hospital, Adelaide, SA, AustraliaDepartment of Neurosciences, Box Hill Hospital, Eastern Health Clinical School, Monash University, Melbourne, VIC, AustraliaDepartment of Neurology, Huashan Hospital, Fudan University, Shanghai, ChinaDepartment of Neurology, Huashan Hospital, Fudan University, Shanghai, ChinaDivision of Neuroradiology, Department of Radiology, University of Ottawa and The Ottawa Hospital, ON, CanadaDepartment of Neurology and Neurophysiology, Liverpool Hospital, Sydney, NSW, Australia; South Western Sydney Clinical School, University of New South Wales, Sydney, NSW, Australia; Ingham Institute for Applied Medical Research, Sydney, NSW, AustraliaBackground: Posterior circulation infarction (POCI) is common. Imaging techniques such as non-contrast-CT (NCCT) and diffusion-weighted-magnetic-resonance-imaging commonly fail to detect hyperacute POCI. Studies suggest expert inspection of Computed Tomography Perfusion (CTP) improves diagnosis of POCI. In many settings, there is limited access to specialist expertise. Deep-learning has been successfully applied to automate imaging interpretation. This study aimed to develop and validate a deep-learning approach for the classification of POCI using CTP. Methods: Data were analysed from 3541-patients from the International-stroke-perfusion-registry (INSPIRE). All patients with baseline multimodal-CT and follow-up imaging performed at 24–48 h were identified. A cohort of 541-patients was constructed on a 1:3 POCI-to −reference-ratio for model analysis. A 3D-Dense-Convolutional-Network (DenseNet) was trained to classify patients into POCI or non-POCI using CTP-deconvolved-maps. Six-stroke-experts also independently classified patients based upon stepwise access to multimodal CT (mCT) data. DenseNet results were compared against expert clinician results. Model and clinician performance was evaluated using area-under-the-receiver-operating-curve, sensitivity, specificity, accuracy and precision. Clinician agreement was measured with the Fleiss-Kappa-statistic. Results: Best mean clinician diagnostic accuracy, sensitivity and agreement was demonstrated after review of all mCT data (AUC: 0.81, Sensitivity: 0.65, Fleiss-Kappa-statistic: 0.73). There was a spectrum of individual clinician results with an AUC-range of 0.73–0.86. Best DenseNet performance was recorded with an input combination of NCCT and delay-time maps. The DenseNet model was superior to the best mean clinician performance (AUC: 0.87) and was due to enhanced sensitivity (DenseNET: 0.77, Clinician: 0.65). The degree to which the DenseNet model outperformed each clinician ranged and was clinician specific (AUC improvement 0.01–0.14). Conclusion: Comprehensive review of CTP improves diagnostic performance and agreement amongst clinicians. A DenseNet model was superior to best mean clinician performance. The degree of improvement varied by specific clinician. Development of a clinician-DenseNet approach may improve inter-clinician agreement and diagnostic accuracy. This approach may alleviate limited specialist services in resource constrained settings.http://www.sciencedirect.com/science/article/pii/S2213158225000026Ischaemic strokeCT perfusionDeep learningPosterior circulation stroke |
| spellingShingle | Leon S. Edwards Milanka Visser Cecilia Cappelen-Smith Dennis Cordato Andrew Bivard Leonid Churilov Christopher Blair James Thomas Angela Dos Santos Longting Lin Chushuang Chen Carlos Garcia-Esperon Kenneth Butcher Tim Kleinig Phillip MC Choi Xin Cheng Qiang Dong Richard I. Aviv Mark W. Parsons A deep learning approach versus expert clinician panel in the classification of posterior circulation infarction NeuroImage: Clinical Ischaemic stroke CT perfusion Deep learning Posterior circulation stroke |
| title | A deep learning approach versus expert clinician panel in the classification of posterior circulation infarction |
| title_full | A deep learning approach versus expert clinician panel in the classification of posterior circulation infarction |
| title_fullStr | A deep learning approach versus expert clinician panel in the classification of posterior circulation infarction |
| title_full_unstemmed | A deep learning approach versus expert clinician panel in the classification of posterior circulation infarction |
| title_short | A deep learning approach versus expert clinician panel in the classification of posterior circulation infarction |
| title_sort | deep learning approach versus expert clinician panel in the classification of posterior circulation infarction |
| topic | Ischaemic stroke CT perfusion Deep learning Posterior circulation stroke |
| url | http://www.sciencedirect.com/science/article/pii/S2213158225000026 |
| work_keys_str_mv | AT leonsedwards adeeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT milankavisser adeeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT ceciliacappelensmith adeeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT denniscordato adeeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT andrewbivard adeeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT leonidchurilov adeeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT christopherblair adeeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT jamesthomas adeeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT angeladossantos adeeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT longtinglin adeeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT chushuangchen adeeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT carlosgarciaesperon adeeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT kennethbutcher adeeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT timkleinig adeeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT phillipmcchoi adeeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT xincheng adeeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT qiangdong adeeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT richardiaviv adeeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT markwparsons adeeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT leonsedwards deeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT milankavisser deeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT ceciliacappelensmith deeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT denniscordato deeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT andrewbivard deeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT leonidchurilov deeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT christopherblair deeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT jamesthomas deeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT angeladossantos deeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT longtinglin deeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT chushuangchen deeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT carlosgarciaesperon deeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT kennethbutcher deeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT timkleinig deeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT phillipmcchoi deeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT xincheng deeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT qiangdong deeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT richardiaviv deeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction AT markwparsons deeplearningapproachversusexpertclinicianpanelintheclassificationofposteriorcirculationinfarction |