FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W4207013230> ?p ?o ?g. }

• W4207013230 endingPage "443" @default.
• W4207013230 startingPage "442" @default.
• W4207013230 abstract "HomeRadiologyVol. 303, No. 2 PreviousNext Reviews and CommentaryFree AccessEditorialPostprocedural Pneumothorax Detection by Deep Learning on Chest RadiographsMark L. Schiebler , Michael HartungMark L. Schiebler , Michael HartungAuthor AffiliationsFrom the Department of Radiology, UW–Madison School of Medicine and Public Health, 600 Highland Ave, E3/378 Clinical Science Center, Madison, WI 53794.Address correspondence to M.L.S. (e-mail: [email protected]).Mark L. Schiebler Michael HartungPublished Online:Jan 25 2022https://doi.org/10.1148/radiol.212973MoreSectionsPDF ToolsImage ViewerAdd to favoritesCiteTrack CitationsPermissionsReprints ShareShare onFacebookTwitterLinked In See also the article by Hong et al in this issueMark L. Schiebler, MD, is a professor of cardiothoracic radiology at the University of Wisconsin School of Medicine and Public Health with interests in the CT and MR imaging of asthma, atherosclerotic cardiovascular disease, and the metabolic consequences of low muscle density. He currently also serves as the deputy editor of thoracic imaging for Radiology and is a member of the Fleischner Society, the Society for Thoracic Radiology, the ISMRM, and the RSNA.Download as PowerPointOpen in Image Viewer Michael Hartung, MD, is an associate professor of abdominal radiology at the University of Wisconsin School of Medicine and Public Health with interests in educational innovation through technology, radiology reporting, and international radiology service and education. He runs an educational website for abdominal radiology entitled LearnAbdominal.com.Download as PowerPointOpen in Image Viewer Chest radiography remains the most frequently used first modality for the identification of pneumothorax because it is easy, low-cost, rapid, and noninvasive. It is of critical importance in the setting of a tension pneumothorax, which, if not aggressively decompressed, may result in death. In clinical departments performing percutaneous image-guided chest interventions such as biopsy or ablation, a postprocedural upright chest radiograph is obtained to determine if there is a pneumothorax, generally within 1 hour (1). If present, a sizeable, enlarging, and/or symptomatic pneumothorax usually warrants further intervention with salvage techniques (eg, catheter evacuation with pleural blood patch) or the placement of a chest tube (1,2). Small, asymptomatic, and stable (established with a follow-up chest radiograph 2 hours later) pneumothoraces usually do not require any further intervention. Other complications (eg, hemopneumothorax, tension pneumothorax, pneumomediastinum) can also be evaluated at this follow-up examination.The incidence of pneumothorax after percutaneous biopsy has a published range of 10%–45%. The rate of intervention with a chest tube and hospitalization is 0.7%–16% (3,4). The mortality from percutaneous chest interventions is relatively low (0.02%–0.15%) (5,6) but is not entirely without the risk of death. Although chest radiography is frequently used as a first modality to identify pneumothorax, it is less sensitive than CT, the reference standard. Recently, Ianniello et al (7) have shown that bedside chest US in the emergency department is also a very sensitive test for the detection of pneumothorax after chest trauma. Chest radiography is an inexpensive and readily available method for follow-up after chest interventional procedures, despite its lower sensitivity than CT or chest US. Thus, improving the sensitivity of chest radiography for the identification of pneumothorax is an important aim.To address this challenge, in this issue of Radiology, Hong and colleagues (8) demonstrate the utility of a deep learning (DL) method for the identification of pneumothorax after percutaneous biopsy of pulmonary lesions. Recently, Thian et al (9) showed the utility of a DL method for the diagnosis of pneumothorax. They found that the area under the curve values for pneumothorax detection from their six external test sets ranged from 0.91 to 0.98 (9). In the study by Hong et al, the authors studied the issue of pneumothorax detection in two groups of patients who had undergone percutaneous chest intervention. A DL system for pneumothorax detection was used in one group; radiographs in the other group were interpreted without DL system assistance. The authors had an excellent design for their method before they began their research.Their purpose was to determine if a DL method could improve the detection of pneumothorax on chest radiographs after a chest intervention. They also calculated the number of patients (power calculation) that would be needed to perform their study, assuming that the sensitivity using DL would be 20% better than the current standard of radiology reading. As outcomes are important for medical imaging interventions, the authors made the a priori determination to choose the pooled efficacy between their two groups as their primary outcome (8). Secondary aims of their study included comparing the following differences between their two groups: (a) pneumothorax amount, (b) the time interval from follow-up chest radiography to drainage catheter insertion, and (c) the drainage catheter dwelling time.Hong et al then compared the results of the two groups. The first group underwent chest biopsy followed by chest radiography with DL detection and radiologist interpretation, while the second group underwent the routine workflow of radiologist interpretation (no DL). In their retrospective analysis of 1319 patients, they showed that the efficacy of their DL method was superior to radiologists’ interpretation of pneumothorax detection on chest radiographs after needle biopsy (sensitivity, 67.1% vs 85.4%; negative predictive value, 91.3% vs 96.8%; and accuracy, 92.3% vs 96.8%) (8). They also showed that their DL system improved the detection sensitivity for a small amount of pneumothoraces (15% or smaller). These results are not particularly surprising.The authors also found that using the DL system changed outcome. Fewer patients with pneumothorax required a subsequent drainage catheter when the DL system was applied (2% vs 5%; P = .009). In addition, 90 patients (3%) required subsequent drainage catheter insertion (74 [3%] in the non-DL group and 16 [2%] in the DL-applied group) (1). This was an unexpected result, and we need to ask why this occurred. In exploring their results regarding this improved outcome in patients after biopsy, they used the metric of timeliness to intervention for chest tube placement between the two groups (those with DL and those without DL). The key ingredient for the improved performance of the DL group versus the non-DL group was time to identification of the pneumothorax; those chest radiographs with DL were more quickly identified as having a pneumothorax. Therefore, those with a small pneumothorax could be followed up at a sooner interval to determine if it is growing and managed appropriately with evacuation and/or blood patch or chest tube.What do these results mean for imaging departments? Is having a DL system for pneumothorax detection really a mission-critical issue? Is patient care compromised without this specific application of DL to chest radiographs? While these are all good questions, this avoids the proverbial elephant in the room: Can DL be relied on to perform chest radiograph interpretation as a final read? The answer may depend on the situation. During space travel, there will be no radiologists for image analysis of injured astronauts; a DL-based image interpretation system will need to be used for all manner of imaging-related diagnoses, including final diagnoses. Although the example of space travel is extreme, there are multiple other Earth-based situations where radiologists are not immediately available. Thus, one answer to the question of DL diagnosis and its role as a mission-critical service is, “Yes.” While the legal conventions for DL medical imaging and final read interpretations have not yet been passed into law on Earth (or for deep space), the results of this study show that a DL system can serve as an acceptable surrogate to a final read when no imaging specialist is available.How medicine works with the promise of DL methodology for the diagnosis and treatment of disease will remain a major topic of the next few decades. This technology will likely help to facilitate the delivery of higher quality medical care at a lower cost to the consumer.Disclosures of Conflicts of Interest: M.L.S. Shareholder in Stemina Biomarker Discovery, X-Vax, Healthmyne, Elucent Medical, and Elucent Oncology; member of the Radiology editorial board. M.H. No relevant relationships.References1. Zlevor AM, Mauch SC, Knott EA, et al. Percutaneous Lung Biopsy with Pleural and Parenchymal Blood Patching: Results and Complications from 1,112 Core Biopsies. J Vasc Interv Radiol 2021;32(9):1319–1327. Crossref, Medline, Google Scholar2. Wagner JM, Hinshaw JL, Lubner MG, et al. CT-guided lung biopsies: pleural blood patching reduces the rate of chest tube placement for postbiopsy pneumothorax. AJR Am J Roentgenol 2011;197(4):783–788. Crossref, Medline, Google Scholar3. Hiraki T, Mimura H, Gobara H, et al. Incidence of and risk factors for pneumothorax and chest tube placement after CT fluoroscopy-guided percutaneous lung biopsy: retrospective analysis of the procedures conducted over a 9-year period. AJR Am J Roentgenol 2010;194(3):809–814. Crossref, Medline, Google Scholar4. Yoon SH, Park CM, Lee KH, et al. Analysis of Complications of Percutaneous Transthoracic Needle Biopsy Using CT-Guidance Modalities In a Multicenter Cohort of 10568 Biopsies. Korean J Radiol 2019;20(2):323–33.[Published correction appears in Korean J Radiol 2019;20(3):531.]. Crossref, Medline, Google Scholar5. Tomiyama N, Yasuhara Y, Nakajima Y, et al. CT-guided needle biopsy of lung lesions: a survey of severe complication based on 9783 biopsies in Japan. Eur J Radiol 2006;59(1):60–64. Crossref, Medline, Google Scholar6. Richardson CM, Pointon KS, Manhire AR, Macfarlane JT. Percutaneous lung biopsies: a survey of UK practice based on 5444 biopsies. Br J Radiol 2002;75(897):731–735. Crossref, Medline, Google Scholar7. Ianniello S, Piccolo CL, Trinci M, Ajmone Cat CA, Miele V. Extended-FAST plus MDCT in pneumothorax diagnosis of major trauma: time to revisit ATLS imaging approach? J Ultrasound 2019;22(4):461–469. Crossref, Medline, Google Scholar8. Hong W, Hwang EJ, Lee JH, et al. Deep Learning for Detecting Pneumothorax on Chest Radiographs after Needle Biopsy: Clinical Implementation. Radiology 2022;303(2):433–441. Abstract, Google Scholar9. Thian YL, Ng D, Hallinan JTPD, et al. Deep Learning Systems for Pneumothorax Detection on Chest Radiographs: A Multicenter External Validation Study. Radiol Artif Intell 2021;3(4):e200190. Link, Google ScholarArticle HistoryReceived: Nov 22 2021Revision requested: Nov 29 2021Revision received: Dec 1 2021Accepted: Dec 8 2021Published online: Jan 25 2022Published in print: May 2022 FiguresReferencesRelatedDetailsAccompanying This ArticleDeep Learning for Detecting Pneumothorax on Chest Radiographs after Needle Biopsy: Clinical ImplementationJan 25 2022RadiologyRecommended Articles Deep Learning for Detecting Pneumothorax on Chest Radiographs after Needle Biopsy: Clinical ImplementationRadiology2022Volume: 303Issue: 2pp. 433-441CT-guided Lung Biopsy: Effect of Biopsy-side Down Position on Pneumothorax and Chest Tube PlacementRadiology2019Volume: 292Issue: 1pp. 190-196Autologous Blood Patch Injection versus Hydrogel Plug in CT-guided Lung Biopsy: A Prospective Randomized TrialRadiology2018Volume: 290Issue: 2pp. 547-554The PEARL Approach for CT-guided Lung Biopsy: Assessment of Complication RateRadiology2021Volume: 302Issue: 2pp. 473-480Diagnosis and Management of Complications from Percutaneous Biliary Tract InterventionsRadioGraphics2017Volume: 37Issue: 2pp. 665-680See More RSNA Education Exhibits Pleural Effusion: What Every Radiologist Should KnowDigital Posters2021Pop Under Pressure - Common and Uncommon Imaging Findings in BarotraumaDigital Posters2022How Good Are You at Tubes and Lines? Â Digital Posters2019 RSNA Case Collection Pneumothorax Ex VacuoRSNA Case Collection2022Pulmonary aspergillomaRSNA Case Collection2020Pulmonary metastasis causing spontaneous pneumothoraxRSNA Case Collection2020 Vol. 303, No. 2 Metrics Altmetric Score PDF download" @default.
• W4207013230 created "2022-01-26" @default.
• W4207013230 creator A5023251263 @default.
• W4207013230 creator A5027690302 @default.
• W4207013230 date "2022-05-01" @default.
• W4207013230 modified "2023-09-28" @default.
• W4207013230 title "Postprocedural Pneumothorax Detection by Deep Learning on Chest Radiographs" @default.
• W4207013230 cites W1983267785 @default.
• W4207013230 cites W2026900338 @default.
• W4207013230 cites W2100799359 @default.
• W4207013230 cites W2129952408 @default.
• W4207013230 cites W2911005357 @default.
• W4207013230 cites W2988907851 @default.
• W4207013230 cites W3153714520 @default.
• W4207013230 cites W3180702133 @default.
• W4207013230 cites W4206936430 @default.
• W4207013230 doi "https://doi.org/10.1148/radiol.212973" @default.
• W4207013230 hasPubMedId "https://pubmed.ncbi.nlm.nih.gov/35076307" @default.
• W4207013230 hasPublicationYear "2022" @default.
• W4207013230 type Work @default.
• W4207013230 citedByCount "0" @default.
• W4207013230 crossrefType "journal-article" @default.
• W4207013230 hasAuthorship W4207013230A5023251263 @default.
• W4207013230 hasAuthorship W4207013230A5027690302 @default.
• W4207013230 hasConcept C126838900 @default.
• W4207013230 hasConcept C2778329176 @default.
• W4207013230 hasConcept C36454342 @default.
• W4207013230 hasConcept C71924100 @default.
• W4207013230 hasConceptScore W4207013230C126838900 @default.
• W4207013230 hasConceptScore W4207013230C2778329176 @default.
• W4207013230 hasConceptScore W4207013230C36454342 @default.
• W4207013230 hasConceptScore W4207013230C71924100 @default.
• W4207013230 hasIssue "2" @default.
• W4207013230 hasLocation W42070132301 @default.
• W4207013230 hasLocation W42070132302 @default.
• W4207013230 hasOpenAccess W4207013230 @default.
• W4207013230 hasPrimaryLocation W42070132301 @default.
• W4207013230 hasRelatedWork W2048640854 @default.
• W4207013230 hasRelatedWork W2334188953 @default.
• W4207013230 hasRelatedWork W2350816062 @default.
• W4207013230 hasRelatedWork W2352464164 @default.
• W4207013230 hasRelatedWork W2361678546 @default.
• W4207013230 hasRelatedWork W2370758048 @default.
• W4207013230 hasRelatedWork W2382792745 @default.
• W4207013230 hasRelatedWork W2383766656 @default.
• W4207013230 hasRelatedWork W2440004924 @default.
• W4207013230 hasRelatedWork W4241455997 @default.
• W4207013230 hasVolume "303" @default.
• W4207013230 isParatext "false" @default.
• W4207013230 isRetracted "false" @default.
• W4207013230 workType "article" @default.