Back

Contusion Volume is a Cross-cohort Predictor of Delayed Seizures after Traumatic Brain Injury

Nanda, A.; Sun, X.; Schaper, F. L. W. V. J.; Kim, J. A.; Shi, H.; Cohen-Zimerman, S.; Markowitz, A. J.; Rosenthal, E. S.; Fox, M. D.; Edlow, B. L.; Grafman, J. H.; Manley, G. T.; Giacino, J. T.; Jain, S.; Bodien, Y. G.; Snider, S. B.

2026-06-02 neurology
10.64898/2026.06.01.26354621 medRxiv
Show abstract

Objective: Predicting specific cognitive, psychiatric, and health-related sequelae in patients after acute traumatic brain injury (TBI) remains an important but challenging clinical problem. Acute phase computed tomography (CT) scans acquired show hemorrhagic contusions, a common type of traumatic pathology. However, whether CT-measured contusions predict long-term sequelae is uncertain. Methods: We established a Screening Cohort of patients with acute TBI who received care at a single TBI Model Systems (TBIMS) inpatient rehabilitation facility. Regions of hemorrhagic contusion and edema were labeled on acute brain CT scans using the fully-automated Brain Lesion Analysis and Segmentation Tool (BLAST-CT). We screened 198 outcome variables at 1-year post-injury for association with acute hemorrhagic contusion volume using the Harrell's Concordance index (C-index), controlling for multiple comparisons using 5,000 outcome permutations. Finally, we tested whether the significant associations in the TBIMS database replicated in acute (Transforming Research and Clinical Knowledge in TBI [TRACK-TBI]) and chronic (Vietnam Head Injury Study [VHIS]) external validation cohorts. Results: The TBIMS Screening Cohort included 345 participants (mean {+/-} SD age: 55.7 {+/-} 21.5 years) with median [IQR] contusion volume 2.3 cc [0.1, 14.6]. Among 198 candidate outcome variables, only delayed seizures were significantly associated with acute hemorrhagic contusion volume (C-index = 0.81; PFWE = 0.007). Contusion volume was not significantly associated with commonly-used measures of global functioning like the Glasgow Outcome Scale Extended, (C-index = 0.55; PFWE = 1). Within the screening cohort, 30 ccs was the optimal volume threshold for discriminating patients with versus without delayed seizures (OR 12.6, 95% CI: [4.6, 34.3]). Contusions larger than 30 cc remained significantly associated with delayed seizures in two external cohorts: (TRACK-TBI OR 4.1 [1.5, 11.2]; VHIS OR 3.2 [1.7, 6.2]). Interpretation: Across three cohorts of patients with TBI, CT-derived contusion volume is robustly associated with the development of delayed seizures, in contrast to commonly-used outcomes measuring global functioning. A 30-cc volume threshold can be used to improve epilepsy prediction models and enrich populations for clinical trials.

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.