Literature DB >> 33122182

Reliability Analysis of Deep Learning Algorithms for Reporting of Routine Lumbar MRI Scans.

Kai-Uwe Lewandrowski1, Narendran Muraleedharan2, Steven Allen Eddy3, Vikram Sobti4, Brian D Reece5, Jorge Felipe Ramírez León6, Sandeep Shah3.   

Abstract

BACKGROUND: Artificial intelligence could provide more accurate magnetic resonance imaging (MRI) predictors of successful clinical outcomes in targeted spine care.
OBJECTIVE: To analyze the level of agreement between lumbar MRI reports created by a deep learning neural network (RadBot) and the radiologists' MRI reading.
METHODS: The compressive pathology definitions were extracted from the radiologist lumbar MRI reports from 65 patients with a total of 383 levels for the central canal: (0) no disc bulge/protrusion/canal stenosis, (1) disc bulge without canal stenosis, (2) disc bulge resulting in canal stenosis, and (3) disc herniation/protrusion/extrusion resulting in canal stenosis. For both, neural foramina were assessed with either (0) neural foraminal stenosis absent or (1) neural foramina stenosis present. Reporting criteria for the pathologies at each disc level and, when available, the grading of severity were extracted, and the Natural Language Processing model was used to generate a verbal and written report. The RadBot report was analyzed similarly as the MRI report by the radiologist. MRI reports were investigated by dichotomizing the data into 2 categories: normal and stenosis. The quality of the RadBot test was assessed by determining its sensitivity, specificity, and positive and negative predictive value as well as its reliability with the calculation of the Cronbach alpha and Cohen kappa using the radiologist MRI report as a gold standard.
RESULTS: The authors found a RadBot sensitivity of 73.3%, a specificity of 88.4%, a positive predictive value of 80.3%, and a negative predictive value of 83.7%. The reliability analysis revealed the Cronbach alpha as 0.772. The highest individual values of the Cronbach alpha were 0.629 and 0.681 when compared to the MRI report by the radiologist, rending values of 0.566 and 0.688, respectively. Analysis of interobserver reliability rendered an overall kappa for the RadBot of 0.627. Analysis of receiver operating characteristics (ROC) showed a value of 0.808 for the area under the ROC curve.
CONCLUSIONS: Deep learning algorithms, when used for routine reporting in lumbar spine MRI, showed excellent quality as a diagnostic test that can distinguish the presence of neural element compression (stenosis) at a statistically significant level (P < .0001) from a random event distribution. This research should be extended to validated and directly visualized pain generators to improve the accuracy and prognostic value of the routine lumbar MRI scan for favorable clinical outcomes with intervention and surgery. LEVEL OF EVIDENCE: 3. CLINICAL RELEVANCE: Validity, clinical teaching, and evaluation study. This manuscript is generously published free of charge by ISASS, the International Society for the Advancement of Spine Surgery.
Copyright © 2020 ISASS.

Entities:  

Keywords:  artificial intelligence; deep neural network learning; magnetic resonance imaging; reliability analysis; spinal pathologies

Year:  2020        PMID: 33122182      PMCID: PMC7735462          DOI: 10.14444/7132

Source DB:  PubMed          Journal:  Int J Spine Surg        ISSN: 2211-4599


  45 in total

1.  Hindsight bias, outcome knowledge and adaptive learning.

Authors:  K Henriksen; H Kaplan
Journal:  Qual Saf Health Care       Date:  2003-12

2.  A practical MRI grading system for lumbar foraminal stenosis.

Authors:  Seunghun Lee; Joon Woo Lee; Jin Sup Yeom; Ki-Jeong Kim; Hyun-Jib Kim; Soo Kyo Chung; Heung Sik Kang
Journal:  AJR Am J Roentgenol       Date:  2010-04       Impact factor: 3.959

3.  Epidural corticosteroid injections for sciatica: placebo effect, injection effect or anti-inflammatory effect?

Authors:  Jean-Pierre Valat
Journal:  Nat Clin Pract Rheumatol       Date:  2006-10

4.  A new grading system of lumbar central canal stenosis on MRI: an easy and reliable method.

Authors:  Guen Young Lee; Young Lee Guen; Joon Woo Lee; Woo Lee Joon; Hee Seok Choi; Seok Choi Hee; Kyoung-Jin Oh; Oh Kyoung-Jin; Heung Sik Kang; Sik Kang Heung
Journal:  Skeletal Radiol       Date:  2011-02-01       Impact factor: 2.199

5.  Re: Diagnostic tests the NASS stenosis guidelines.

Authors:  D Scott Kreiner; Jamie Baisden; Thomas Gilbert; William O Shaffer; Jeffrey T Summers
Journal:  Spine J       Date:  2013-10-09       Impact factor: 4.166

6.  Supervised methods for detection and segmentation of tissues in clinical lumbar MRI.

Authors:  Subarna Ghosh; Vipin Chaudhary
Journal:  Comput Med Imaging Graph       Date:  2014-03-31       Impact factor: 4.790

7.  Impact of occupational characteristics on return to work for employed patients after elective lumbar spine surgery.

Authors:  Inamullah Khan; Mohamad Bydon; Kristin R Archer; Ahilan Sivaganesan; Anthony M Asher; Muhammad Ali Alvi; Panagiotis Kerezoudis; John J Knightly; Kevin T Foley; Erica F Bisson; Christopher Shaffrey; Anthony L Asher; Dan M Spengler; Clinton J Devin
Journal:  Spine J       Date:  2019-08-20       Impact factor: 4.166

8.  Diagnostic accuracy of low-dose versus ultra-low-dose CT for lumbar disc disease and facet joint osteoarthritis in patients with low back pain with MRI correlation.

Authors:  Sun Hwa Lee; Seong Jong Yun; Hyeon Hwan Jo; Dong Hyeon Kim; Jae Gwang Song; Yong Sung Park
Journal:  Skeletal Radiol       Date:  2017-11-06       Impact factor: 2.199

9.  The prognosis of acute low back pain in primary care in the United States: a 2-year prospective cohort study.

Authors:  Wolf E Mehling; Viranjini Gopisetty; Elizabeth Bartmess; Mike Acree; Alice Pressman; Harley Goldberg; Frederick M Hecht; Tim Carey; Andrew L Avins
Journal:  Spine (Phila Pa 1976)       Date:  2012-04-15       Impact factor: 3.468

10.  Targeted methylprednisolone acetate/hyaluronidase/clonidine injection after diagnostic epiduroscopy for chronic sciatica: a prospective, 1-year follow-up study.

Authors:  Jos W Geurts; Jan-Willem Kallewaard; Jonathan Richardson; Gerbrand J Groen
Journal:  Reg Anesth Pain Med       Date:  2002 Jul-Aug       Impact factor: 6.288

View more
  2 in total

1.  A deep learning model for detection of cervical spinal cord compression in MRI scans.

Authors:  Zamir Merali; Justin Z Wang; Jetan H Badhiwala; Christopher D Witiw; Jefferson R Wilson; Michael G Fehlings
Journal:  Sci Rep       Date:  2021-05-18       Impact factor: 4.379

2.  A Proposed Personalized Spine Care Protocol (SpineScreen) to Treat Visualized Pain Generators: An Illustrative Study Comparing Clinical Outcomes and Postoperative Reoperations between Targeted Endoscopic Lumbar Decompression Surgery, Minimally Invasive TLIF and Open Laminectomy.

Authors:  Kai-Uwe Lewandrowski; Ivo Abraham; Jorge Felipe Ramírez León; Albert E Telfeian; Morgan P Lorio; Stefan Hellinger; Martin Knight; Paulo Sérgio Teixeira De Carvalho; Max Rogério Freitas Ramos; Álvaro Dowling; Manuel Rodriguez Garcia; Fauziyya Muhammad; Namath Hussain; Vicky Yamamoto; Babak Kateb; Anthony Yeung
Journal:  J Pers Med       Date:  2022-06-29
  2 in total

北京卡尤迪生物科技股份有限公司 © 2022-2023.