Performance of a 3D convolutional neural network in the detection of hypoperfusion at CT pulmonary angiography in patients with chronic pulmonary embolism : a feasibility study

Show simple item record

dc.contributor.author Vainio, Tuomas J V
dc.contributor.author Mäkelä, Teemu Olavi
dc.contributor.author Savolainen, Sauli
dc.contributor.author Kangasniemi, Marko Matti
dc.date.accessioned 2021-09-29T05:38:03Z
dc.date.available 2021-09-29T05:38:03Z
dc.date.issued 2021-09-24
dc.identifier.citation Vainio , T J V , Mäkelä , T O , Savolainen , S & Kangasniemi , M M 2021 , ' Performance of a 3D convolutional neural network in the detection of hypoperfusion at CT pulmonary angiography in patients with chronic pulmonary embolism : a feasibility study ' , European radiology experimental , vol. 5 , no. 1 , 45 . https://doi.org/10.1186/s41747-021-00235-z
dc.identifier.other PURE: 168818307
dc.identifier.other PURE UUID: 20e48d65-f6ff-48fb-b945-3af119a61859
dc.identifier.other ORCID: /0000-0001-8085-322X/work/100759299
dc.identifier.other WOS: 000698670400001
dc.identifier.uri http://hdl.handle.net/10138/334685
dc.description.abstract Background Chronic pulmonary embolism (CPE) is a life-threatening disease easily misdiagnosed on computed tomography. We investigated a three-dimensional convolutional neural network (CNN) algorithm for detecting hypoperfusion in CPE from computed tomography pulmonary angiography (CTPA). Methods Preoperative CTPA of 25 patients with CPE and 25 without pulmonary embolism were selected. We applied a 48%-12%-40% training-validation-testing split (12 positive and 12 negative CTPA volumes for training, 3 positives and 3 negatives for validation, 10 positives and 10 negatives for testing). The median number of axial images per CTPA was 335 (min-max, 111-570). Expert manual segmentations were used as training and testing targets. The CNN output was compared to a method in which a Hounsfield unit (HU) threshold was used to detect hypoperfusion. Receiver operating characteristic area under the curve (AUC) and Matthew correlation coefficient (MCC) were calculated with their 95% confidence interval (CI). Results The predicted segmentations of CNN showed AUC 0.87 (95% CI 0.82-0.91), those of HU-threshold method 0.79 (95% CI 0.74-0.84). The optimal global threshold values were CNN output probability >= 0.37 and en
dc.format.extent 12
dc.language.iso eng
dc.relation.ispartof European radiology experimental
dc.rights cc_by
dc.rights.uri info:eu-repo/semantics/openAccess
dc.subject ATTENUATION
dc.subject COMPUTED-TOMOGRAPHY
dc.subject Computed tomography angiography
dc.subject DIAGNOSIS
dc.subject DISEASE
dc.subject Deep learning
dc.subject Feasibility Studies
dc.subject HYPERTENSION
dc.subject Neural networks (computer)
dc.subject PERFUSION
dc.subject PLATFORM
dc.subject Pulmonary embolism
dc.subject THROMBOEMBOLISM
dc.subject 114 Physical sciences
dc.title Performance of a 3D convolutional neural network in the detection of hypoperfusion at CT pulmonary angiography in patients with chronic pulmonary embolism : a feasibility study en
dc.type Article
dc.contributor.organization Doctoral Programme in Clinical Research
dc.contributor.organization HUS Medical Imaging Center
dc.contributor.organization Doctoral Programme in Materials Research and Nanosciences
dc.contributor.organization Department of Diagnostics and Therapeutics
dc.contributor.organization Department of Physics
dc.contributor.organization Helsinki In Vivo Animal Imaging Platform (HAIP)
dc.contributor.organization University Management
dc.contributor.organization Sauli Savolainen / Principal Investigator
dc.contributor.organization Clinicum
dc.description.reviewstatus Peer reviewed
dc.relation.doi https://doi.org/10.1186/s41747-021-00235-z
dc.relation.issn 2509-9280
dc.rights.accesslevel openAccess
dc.type.version publishedVersion

Files in this item

Total number of downloads: Loading...

Files Size Format View
s41747_021_00235_z.pdf 4.166Mb PDF View/Open

This item appears in the following Collection(s)

Show simple item record