Related Experiment Video
Updated: Jun 19, 2025
Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images
Published on: April 13, 2013
Clinical prediction scale approach derived from a retrospective study to reduce the number of urgent, low-value
Juana María Plasencia-Martínez1, Elena Otón-González2, Marta Sánchez-Canales2
1Department of Radiology, Hospital General Universitario Morales Meseguer, Murcia, Spain. Plasen79@gmail.com.
Purpose:
Fifty percent of cranial CT scans performed achieve no benefit and entail risks. Our aim is to determine the yield of non-traumatic urgent cranial-CT and develop a pretest clinical probability scale approach.
Methods:
Adult patients seen in our emergency department between 2017-2021 and referred for urgent cranial-CT for non-traumatic reasons were retrospectively recruited and randomly selected. Presenting complaint (PC), demographic variables, Relevant radiological findings (RRF) on the urgent cranial-CT and Relevant clinical-radiological findings (RCRF: admission need or RRF detection on the urgent cranial-CT or cranial CT/MRI in the following three months) were recruited.
Results:
We recruited 702 patients, with median age 62 [47-76] years, 363 (51.7%) females. RCRF were observed in 404 (57.55%); of these, 352 (50.1%) required admission. RRF were detected in 190 (27.06%): 36 acute ischemic and 27 acute hemorrhagic lesions, 115 masses, 9 edema, and 27 hydrocephalus. Predictive PC for urgent cranial-CT were motor, speech, sensory deficits, sudden alteration of mental status, epileptic seizure, cognitive impairment, neurological symptoms in cancer patients, acute headache without a prior history and with meningeal signs; nausea, vomiting, or hypertensive crisis; visual deficits, and dizziness. This algorithm provided sensitivity, specificity, positive predictive value, and negative predictive value (NPV, 95%CI in brackets) of 92.1% (89-94.5%), 27.5% (22.5-33.0%), 63.3% (59.2-67.2%), and 71.9% (62.7-80.0%), to diagnose RCRF, and 97.4% (93.4-99.1%), 21.3% (17.8-25.1%), 31.5% (27.7-35.4%), and 95.6% (90.1-98.6%), to diagnose RRF. In patients not requiring admission (n = 350), the NPV for RRF was 98.8% (93.6-100%); the negative likelihood ratio 0.08 (0.01-0.57), and sensitivity remained at 97.8% (82.2-99.9%). Applying it would have avoided performing 85/350 urgent cranial-CT (24.29%). To find one RRF, we would have gone from performing 7.8 (350/45) to 5.9 (265/45) CTs, failing to diagnose 1/45 (2.2%) RRF.
Conclusions:
This proposed clinical scale could potentially decrease 24% of urgent cranial-CT.
