BACKGROUND: Penetrating head injuries (PHIs) are common in combat operations and most have visible wound paths on computed tomography (CT). OBJECTIVE: We assess agreement between an automated trajectory analysis-based assessment of brain injury and manual tracings of encephalomalacia on CT. METHODS: We analyzed 80 head CTs with ballistic PHI from the Institutional Review Board approved Vietnam head injury registry. Anatomic reports were generated from spatial coordinates of projectile entrance and terminal fragment location. These were compared to manual tracings of the regions of encephalomalacia. Dice's similarity coefficients, kappa, sensitivities, and specificities were calculated to assess agreement. Times required for case analysis were also compared. RESULTS: Results show high specificity of anatomic regions identified on CT with semiautomated anatomical estimates and manual tracings of tissue damage. Radiologist's and medical students' anatomic region reports were similar (Kappa 0.8, t-test p < 0.001). Region of probable injury modeling of involved brain structures was sensitive (0.7) and specific (0.9) compared with manually traced structures. Semiautomated analysis was 9-fold faster than manual tracings. CONCLUSION: Our region of probable injury spatial model approximates anatomical regions of encephalomalacia from ballistic PHI with time-saving over manual methods. Results show potential for automated anatomical reporting as an adjunct to current practice of radiologist/neurosurgical review of brain injury by penetrating projectiles. Reprint &
BACKGROUND: Penetrating head injuries (PHIs) are common in combat operations and most have visible wound paths on computed tomography (CT). OBJECTIVE: We assess agreement between an automated trajectory analysis-based assessment of brain injury and manual tracings of encephalomalacia on CT. METHODS: We analyzed 80 head CTs with ballistic PHI from the Institutional Review Board approved Vietnam head injury registry. Anatomic reports were generated from spatial coordinates of projectile entrance and terminal fragment location. These were compared to manual tracings of the regions of encephalomalacia. Dice's similarity coefficients, kappa, sensitivities, and specificities were calculated to assess agreement. Times required for case analysis were also compared. RESULTS: Results show high specificity of anatomic regions identified on CT with semiautomated anatomical estimates and manual tracings of tissue damage. Radiologist's and medical students' anatomic region reports were similar (Kappa 0.8, t-test p < 0.001). Region of probable injury modeling of involved brain structures was sensitive (0.7) and specific (0.9) compared with manually traced structures. Semiautomated analysis was 9-fold faster than manual tracings. CONCLUSION: Our region of probable injury spatial model approximates anatomical regions of encephalomalacia from ballistic PHI with time-saving over manual methods. Results show potential for automated anatomical reporting as an adjunct to current practice of radiologist/neurosurgical review of brain injury by penetrating projectiles. Reprint &
Authors: Richard Dirnhofer; Christian Jackowski; Peter Vock; Kimberlee Potter; Michael J Thali Journal: Radiographics Date: 2006 Sep-Oct Impact factor: 5.333
Authors: Les R Folio; Tatjana V Fischer; Paul J Shogan; Michael I Frew; Pil S Kang; Rolf Bunger; James M Provenzale Journal: Radiology Date: 2011-01-11 Impact factor: 11.105
Authors: Scott D Steenburg; Clint W Sliker; Kathirkamanathan Shanmuganathan; Eliot L Siegel Journal: Radiographics Date: 2010 Jul-Aug Impact factor: 5.333
Authors: Michael Koenigs; Edward D Huey; Vanessa Raymont; Bobby Cheon; Jeffrey Solomon; Eric M Wassermann; Jordan Grafman Journal: Nat Neurosci Date: 2007-12-23 Impact factor: 24.884