RADAR - Radiomics on aSDH: predicting outcome with surface area

Abstract

Background: Acute subdural hematoma is a critical condition, leading to significant morbidity and mortality. Despite advancements in surgical techniques, a portion of patients only show limited clinical improvement post-evacuation. Surgical intervention decisions are critically important, as they can either improve or worsen a patient's condition. Radiomics offers significant potential by extracting complex patterns from digital medical images and transforming them into high-dimensional data that reflect the underlying pathophysiology. By integrating Radiomics with individual patient characteristics, we can develop decision support models. This study aims to analyze radiomic parameters of aSDH to determine whether they support the decision to proceed with urgent surgery or opt for a conservative approach. We hypothesized that surface area could be a significant predictor of neurological outcome such as maintaining independent mobility (mRS ≥ 3) and survival rates.

Methods: This retrospective study involved radiomic analysis according to neurological outcome and survival. Radiomic parameters were measured using 3D Slicer software. Statistical analyses explored correlations, employing AUC-analysis and Kaplan-Meier survival.

Results: Our findings revealed significant correlations between hematoma and surface area with poorer neurological prognosis. Further subgroup analysis showed surface area as a significant predictor for poorer outcomes in patients undergoing craniotomy (p = 0.006 in univariant- and p = 0.020 in multivariant analysis). In the total cohort, among conservatively managed and craniotomy subgroups, survival analysis highlighted an advantageous survival for patients exhibiting smaller surface areas (< 339.50 cm²).

Conclusions: Especially in craniotomy patients, surface area emerged as a possible predictor for neurological outcome and survival.

Keywords: Acute subdural hematoma; Outcome; Radiomics; Surface area.

Fig. 1

Segmentation in 3D slicer. A Unedited CT-scan. B Marking acute subdural hematoma in red and surrounding areas in green in multiple image planes and all three axes. C Semiautomated two-dimensional reconstruction of hematoma. D Rotatable three-dimensional reconstruction of hematoma that is used for calculating wanted parameters

Fig. 2

Receiver operating curve representing sensitivity against 1-specificity. Feret diameter is displayed by the blue, volume by the turquoise, and surface area by the lilac curve. All parameters show an AUC above 0.6

Fig. 3

Kaplan-Meier survival for total cohort according to surface area. The Kaplan-Meier curve depicts the survival for twelve months of all patients with surface area above (lilac curve) versus below (turquoise curve) the cut-off value for surface area. Also shown is the number of patients at risk on the day of discharge, three months, six months and twelve months after discharge representing the patients that are still alive

Fig. 4

Kaplan-Meier survival for patients treated conservatively according to surface area. The Kaplan-Meier curve depicts the survival for twelve months of patients treated conservatively with surface area above (lilac curve) versus below (turquoise curve) the cut-off value for surface area. Also shown is the number of patients at risk on the day of discharge, three months, six months and twelve months after discharge representing the patients that are still alive

Fig. 5

Kaplan-Meier survival for patients undergoing craniotomy according to surface area. The Kaplan Meier curve displays survival for twelve months of patients undergoing craniotomy with surface area above (lilac curve) versus below (turquoise curve) the cut-off value for surface area. Also shown is the number of patients at risk on the day of discharge, three months, six months and twelve months after discharge representing the patients that are still alive

Fig. 6

Interaction between acute subdural hematoma and brain surface. A: astrocyte. aSDH: acute subdural hematoma. BBB: blood-brain barrier. BOX: bilirubin oxidation products. CA-1: carbonic anhydrase I. MG: microglia. MMP9: matrix metalloproteinase-9. OG: oligodendrocytes. PAR1: protease activated receptor 1. ROS: reactive oxygen species. TNF-α: tumor necrosis factor alpha. CA-1 from lysed erythrocytes lowers the blood-brain barrier, making it easier for thrombin to cross. This can subsequently increase glutamate efflux in astrocytes, leading to a lower epileptic threshold. Additionally, thrombin can bind to the protease activated receptor 1 (PAR 1) in oligodendrocytes, stimulating the release of MMP9 and TNF-α, which can have neuro-damaging effects. Heme from lysed erythrocytes is converted into bilirubin by heme oxygenase-1 and biliverdin reductase in oligodendrocytes and microglia, which is later transformed into bilirubin oxidation products (BOX) by reactive oxygen species leading to a release of cytokines and chemokines. This enables the upregulation of endothelial adhesion molecules resulting in an increased leucocyte migration. Iron from the acute subdural hematoma can react with H₂O₂ and generate reactive oxygen species. These can cause harmful effects such as DNA damage and lipid peroxidation in neurons