Abstract
Traumatic brain injury (TBI) is a leading cause of death and disability in young adults. Prognostication is challenging due to its heterogenous nature. Traumatic axonal injury (TAI)
is a common lesion type in moderate and severe TBI, and indirect signs of TAI on MRI are of prognostic importance. However, more knowledge is needed about which specific MRI
features are most relevant to outcome. The overall aim of this PhD project was to investigate the prognostic value of TAI detected on early MRI in both adults and children with TBI.
In the first paper, we segmented TAI lesions detected on MRI obtained in a clinical setting within 6 weeks after injury in 269 patients (aged 8–70 years) with moderate and
severe TBI. The lesion segmentations were registered to a common brain template and combined into lesion frequency distribution maps. The maps were evaluated visually and also
with use of voxel-based lesion–symptom mapping. The highest frequency of TAI was found in the splenium of corpus callosum. The patients with the poorest global outcomes 1 year
after injury had the highest frequencies of TAI, and TAI in the splenium of corpus callosum and in specific parts of the brainstem were associated with worse outcomes. Our findings
showed that location as well as burden of TAI are important in outcome prediction.
In the second study, we aimed to develop improved gradings of TAI on MRI that could predict outcome better than the current standard grading, which is based on
histopathological studies. We documented the location, number, and volume of TAI lesions on early MRI from 463 patients (aged 8–70 years) with mild, moderate and severe TBI.
Based on analyses on which MRI variables were most important to outcome, we proposed a pragmatic clinical Trondheim TAI-MRI grading with bilateral TAI in the pons and bilateral
TAI in the mesencephalon or the thalami as the worst grades (grade 5 and 4 respectively). For severe TBI, the Trondheim TAI-MRI grading predicted outcome more reliably than existing
clinical gradings. We also proposed quantitative models that included volumes of TAI and brain contusions, and these models outperformed clinical gradings in all TBI severities. In
conclusion, more detailed lesion location and inclusion of lesion burden in the grading of TAI on MRI improved outcome prediction.
In the third study, we investigated the impact of TAI on MRI in 56 children (aged 0– 18 years) with moderate and severe TBI. We found a prevalence of TAI similar to that in
adult populations, and that the severity of TAI (the standard TAI-MRI grading as well as the Trondheim TAI-MRI grading) was associated with increased risk of disability both at 1 year
and 5 years after injury. These findings show that early MRI holds important prognostic information also in pediatric TBI and advocates its implementation in future guidelines for the management of this patient group.
Across the three studies, this thesis demonstrates that early MRI can identify signs of TAI across TBI severities and age groups, and that refined TAI-MRI grading systems
improve prognostic accuracy in both adults and children. The increasing use of artificial intelligence (AI) in radiological workflows seems promising for implementing volumetric
variables. Integrating imaging data with clinical information, biomarkers, and other factors important to outcome, into unified AI-based prognostic models may ultimately enhance
clinical decision-making and support more tailored treatment and rehabilitation.