FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2904153148> ?p ?o ?g. }

• W2904153148 endingPage "681" @default.
• W2904153148 startingPage "680" @default.
• W2904153148 abstract "HomeRadiologyVol. 290, No. 3 PreviousNext Reviews and CommentaryFree AccessEditorialWith a Little Help from Machine Learning, Precision Radiology Can Be FeasiblePaul J. Chang Paul J. Chang Author AffiliationsFrom the Department of Radiology, University of Chicago Medicine, 5841 S Maryland Ave, MC2026, Chicago, IL 60637.Address correspondence to the author (e-mail: [email protected]).Paul J. Chang Published Online:Dec 11 2018https://doi.org/10.1148/radiol.2018182557MoreSectionsPDF ToolsImage ViewerAdd to favoritesCiteTrack CitationsPermissionsReprints ShareShare onFacebookTwitterLinked In See also the article by Weston et al in this issue.IntroductionI have a confession to make: there are days when I feel that I have not provided the best possible care to my patients. The number of complex, ever-increasing-in-size imaging data sets on my picture archiving and communication system worklist seemingly grows each day, resulting in less available time to devote to each patient. This guilt is compounded after reading articles from this and other journals establishing evidence that precision radiology (radiomics and other quantitative actionable phenotypic characterization) can be valuable and impactful to patient care: I am mindful that I cannot devote the time or expense necessary to apply these techniques in my normal practice. Asking our 3D/Advanced Visualization laboratory personnel to perform these typically quantitative and time-consuming tasks is not feasible; they are also overworked.To realize the benefits of precision radiology in a practical way, we need some assistance. Machine learning and other advanced information technology might offer the help we need. They provide an automated way to off-load the time-consuming and inefficient manual processes usually required in these advanced techniques.In this issue of Radiology, Weston and colleagues train a deep learning convolutional neural network to perform automated abdominal CT body composition segmentation (1). Their machine learning algorithm met or exceeded the accuracy of expert (and time-consuming) manual segmentation on test examinations. In addition, the machine learning algorithm also demonstrated promising performance when applied to a separate CT data set from patients with hepatocellular carcinoma.Body composition analysis is an excellent example of actionable quantitative phenotypic characterization. It is typical of precision radiology in that it adds value to the management of a wide variety of cardiovascular, oncologic, and surgical outcome conditions (2–9). It also shares the common liability of precision radiology: it is too time-consuming and inefficient to operationalize in most of our practices. I am particularly attracted to applying machine learning to this specific type-of-use case: a potentially clinically useful segmentation that is too time- or resource-intensive to perform in routine clinical practice. I believe these consumable low-hanging fruit applications of artificial intelligence will be very important as augmentation tools in the future. They will allow radiologists to provide added value as well as improved efficiency and reduced variability in everyday practice.The article by Weston et al also demonstrates that this is still early times for machine learning in radiology. The authors show good segmentation agreement (using traditional performance metrics, such as Dice and Jaccard scores) between the deep learning algorithm and manually segmented test images. Yet, the number of training (n = 2430) and test cases (n = 270) are relatively small, especially since the authors used a holdout validation approach. Other—albeit more complex and experimentally time-consuming—validation approaches, such as K-fold cross validation, theoretically may be more appropriate for small data sets (10). The concern of overfitting and ability to generalize machine learning algorithm performance beyond modest test data sets to the general clinical practice is a nontrivial challenge. It is a reality speaking not only to challenges in experimental design, but also to real-world annotated data availability constraints (ie, lack of IT data interoperability, access, semantic normalization, etc). These practical challenges need to be mitigated before artificial intelligence becomes real.The authors attempted to test the generalizability of their deep learning algorithm on a separate data set of patients with hepatocellular carcinoma. While the analysis did not include discrete segmentation of visceral adipose and fat-free tissues, performance was promising (although, predictably a bit worse). The authors make a correct observation: generalizing their algorithm requires additional features to be modeled. This will again require a substantial increase in annotated training and test data sets.Challenges notwithstanding, this study is a good example of how machine learning can and should be applied to radiology. It is not to replace radiologists but to enable us to provide precision radiology in an efficient and scalable manner for our patients.Disclosures of Conflicts of Interest: P.J.C. Activities related to the present article: disclosed no relevant relationships. Activities not related to the present article: disclosed research grants to institution from Philips and personal payment received from EnvoyAI, AIDoc, Inference Analytics, and Bayer for service on medical advisory boards. Other relationships: disclosed no relevant relationships.References1. Weston AD, Korfiatis P, Kline T, et al. Automated abdominal segmentation of CT scans for body composition analysis using deep learning. Radiology 2019;290:669–679. Link, Google Scholar2. Gonzalez MC, Pastore CA, Orlandi SP, Heymsfield SB. Obesity paradox in cancer: new insights provided by body composition. Am J Clin Nutr 2014;99(5):999–1005. Crossref, Medline, Google Scholar3. Aust S, Knogler T, Pils D, et al. Skeletal muscle depletion and markers for cancer cachexia are strong prognostic factors in epithelial ovarian cancer. PLoS One 2015;10(10):e0140403. Crossref, Medline, Google Scholar4. Malietzis G, Currie AC, Athanasiou T, et al. Influence of body composition profile on outcomes following colorectal cancer surgery. Br J Surg 2016;103(5):572–580. Crossref, Medline, Google Scholar5. Jean N, Somers VK, Sochor O, Medina-Inojosa J, Llano EM, Lopez-Jimenez F. Normal-weight obesity: implications for cardiovascular health. Curr Atheroscler Rep 2014;16(12):464. Crossref, Medline, Google Scholar6. Sheetz KH, Waits SA, Terjimanian MN, et al. Cost of major surgery in the sarcopenic patient. J Am Coll Surg 2013;217(5):813–818. Crossref, Medline, Google Scholar7. Sugimoto M, Farnell MB, Nagorney DM, et al. Decreased skeletal muscle volume is a predictive factor for poorer survival in patients undergoing surgical resection for pancreatic ductal adenocarcinoma. J Gastrointest Surg 2018;22(5):831–839. Crossref, Medline, Google Scholar8. Jacquelin-Ravel N, Pichard C. Clinical nutrition, body composition and oncology: a critical literature review of the synergies. Crit Rev Oncol Hematol 2012;84(1):37–46. Crossref, Medline, Google Scholar9. Caan BJ, Cespedes Feliciano EM, Kroenke CH. The importance of body composition in explaining the overweight paradox in cancer-counterpoint. Cancer Res 2018;78(8):1906–1912. Crossref, Medline, Google Scholar10. Cawley G, Talbot N. On overfitting in model selection and subsequent selection bias in performance evaluation. J Mach Learn Res 2010;11:2079–2107. Google ScholarArticle HistoryReceived: Nov 6 2018Revision requested: Nov 7 2018Revision received: Nov 9 2018Accepted: Nov 14 2018Published online: Dec 11 2018Published in print: Mar 2019 FiguresReferencesRelatedDetailsCited ByApplication of artificial intelligence to imaging interpretations in the musculoskeletal area: Where are we? Where are we going?ValérieBousson, NicolasBenoist, PierreGuetat, GrégoireAttané, CécileSalvat, LaetitiaPerronne2023 | Joint Bone Spine, Vol. 90, No. 1Künstliche Intelligenz im GesundheitswesenAnna L.Kauffmann, JasminHennrich, ChristophBuck, TorstenEymann2022Moving Artificial Intelligence from Feasible to Real: Time to Drill for Gas and Build RoadsPaul J. Chang, 3 December 2019 | Radiology, Vol. 294, No. 2Prognostic value of anthropometric measures extracted from whole-body CT using deep learning in patients with non-small-cell lung cancerPaulBlanc-Durand, LucaCampedel, SébastienMule, SimonJegou, AlainLuciani, FrédéricPigneur, EmmanuelItti2020 | European Radiology, Vol. 30, No. 6Accompanying This ArticleAutomated Abdominal Segmentation of CT Scans for Body Composition Analysis Using Deep LearningDec 11 2018RadiologyRecommended Articles Automated Abdominal Segmentation of CT Scans for Body Composition Analysis Using Deep LearningRadiology2018Volume: 290Issue: 3pp. 669-679Deep Learning: A Primer for RadiologistsRadioGraphics2017Volume: 37Issue: 7pp. 2113-2131Fully Automated and Standardized Segmentation of Adipose Tissue Compartments via Deep Learning in 3D Whole-Body MRI of Epidemiologic Cohort StudiesRadiology: Artificial Intelligence2020Volume: 2Issue: 6Current Applications and Future Impact of Machine Learning in RadiologyRadiology2018Volume: 288Issue: 2pp. 318-328A Fully Automated Deep Learning Pipeline for Multi–Vertebral Level Quantification and Characterization of Muscle and Adipose Tissue on Chest CT ScansRadiology: Artificial Intelligence2022Volume: 4Issue: 1See More RSNA Education Exhibits Advanced CT Techniques In the Evaluation of Hepatocellular Carcinoma: Roles of Ultra-high-resolution CT, Dual-energy CT, Contrast Enhancement Boost Technique, and RadiomicsDigital Posters2022Artificial Intelligence in Diagnostic Imaging: Current Applications and Future PerspectiveDigital Posters2019Imaging and Radiomics in Hepatocellular Carcinoma (HCC): Current Update and Future PerspectivesDigital Posters2021 RSNA Case Collection Brown Adipose TissueRSNA Case Collection2021Esophageal Fibrovascular PolypRSNA Case Collection2022Sternalis muscle - normal variantRSNA Case Collection2020 Vol. 290, No. 3 Metrics Altmetric Score PDF download" @default.
• W2904153148 created "2018-12-22" @default.
• W2904153148 creator A5076555423 @default.
• W2904153148 date "2019-03-01" @default.
• W2904153148 modified "2023-10-03" @default.
• W2904153148 title "With a Little Help from Machine Learning, Precision Radiology Can Be Feasible" @default.
• W2904153148 cites W1913776872 @default.
• W2904153148 cites W1973993374 @default.
• W2904153148 cites W1999648662 @default.
• W2904153148 cites W2022025060 @default.
• W2904153148 cites W2105746949 @default.
• W2904153148 cites W2301680230 @default.
• W2904153148 cites W2787431611 @default.
• W2904153148 cites W2797673408 @default.
• W2904153148 cites W2905023912 @default.
• W2904153148 doi "https://doi.org/10.1148/radiol.2018182557" @default.
• W2904153148 hasPubMedId "https://pubmed.ncbi.nlm.nih.gov/30526348" @default.
• W2904153148 hasPublicationYear "2019" @default.
• W2904153148 type Work @default.
• W2904153148 sameAs 2904153148 @default.
• W2904153148 citedByCount "5" @default.
• W2904153148 countsByYear W29041531482020 @default.
• W2904153148 countsByYear W29041531482022 @default.
• W2904153148 countsByYear W29041531482023 @default.
• W2904153148 crossrefType "journal-article" @default.
• W2904153148 hasAuthorship W2904153148A5076555423 @default.
• W2904153148 hasBestOaLocation W29041531481 @default.
• W2904153148 hasConcept C119857082 @default.
• W2904153148 hasConcept C126838900 @default.
• W2904153148 hasConcept C19527891 @default.
• W2904153148 hasConcept C41008148 @default.
• W2904153148 hasConcept C71924100 @default.
• W2904153148 hasConceptScore W2904153148C119857082 @default.
• W2904153148 hasConceptScore W2904153148C126838900 @default.
• W2904153148 hasConceptScore W2904153148C19527891 @default.
• W2904153148 hasConceptScore W2904153148C41008148 @default.
• W2904153148 hasConceptScore W2904153148C71924100 @default.
• W2904153148 hasIssue "3" @default.
• W2904153148 hasLocation W29041531481 @default.
• W2904153148 hasLocation W29041531482 @default.
• W2904153148 hasOpenAccess W2904153148 @default.
• W2904153148 hasPrimaryLocation W29041531481 @default.
• W2904153148 hasRelatedWork W1995515455 @default.
• W2904153148 hasRelatedWork W2049214470 @default.
• W2904153148 hasRelatedWork W2080531066 @default.
• W2904153148 hasRelatedWork W2081340600 @default.
• W2904153148 hasRelatedWork W2482660221 @default.
• W2904153148 hasRelatedWork W2748952813 @default.
• W2904153148 hasRelatedWork W2899084033 @default.
• W2904153148 hasRelatedWork W2902148150 @default.
• W2904153148 hasRelatedWork W3031052312 @default.
• W2904153148 hasRelatedWork W3032375762 @default.
• W2904153148 hasVolume "290" @default.
• W2904153148 isParatext "false" @default.
• W2904153148 isRetracted "false" @default.
• W2904153148 magId "2904153148" @default.
• W2904153148 workType "article" @default.