Refining outcome prediction after traumatic brain injury with machine learning algorithms.

Outcome after traumatic brain injury (TBI) is typically assessed using the Glasgow outcome scale extended (GOSE) with levels from 1 (death) to 8 (upper good recovery). Outcome prediction has classically been dichotomized into either dead/alive or favorable/unfavorable outcome. Binary outcome prediction models limit the possibility of detecting subtle yet significant improvements. We set out to explore different machine learning methods with the purpose of mapping their predictions to the full 8 grade scale GOSE following TBI. The models were set up using the variables: age, GCS-motor score, pupillary reaction, and Marshall CT score. For model setup and internal validation, a total of 866 patients could be included. For external validation, a cohort of 369 patients were included from Leuven, Belgium, and a cohort of 573 patients from the US multi-center ProTECT III study. Our findings indicate that proportional odds logistic regression (POLR), random forest regression, and a neural network model achieved accuracy values of 0.3-0.35 when applied to internal data, compared to the random baseline which is 0.125 for eight categories. The models demonstrated satisfactory performance during external validation in the data from Leuven, however, their performance were not satisfactory when applied to the ProTECT III dataset.

Prognostic scores for survival as decisional support for surgery in spinal metastases: a performance assessment systematic review.

PURPOSE: To review the evidence on the relative prognostic performance of the available prognostic scores for survival in spinal metastatic surgery in order to provide a recommendation for use in clinical practice. METHODS: A systematic review of comparative external validation studies assessing the performance of prognostic scores for survival in independent cohorts was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines. Eligible studies were identified through Medline and Embase until May 2021. Studies were included when they compared at least four survival scoring systems in surgical or mixed cohorts across all primary tumor types. Predictive performance was assessed based on discrimination and calibration for 3-month, 1-year and overall survival, and generalizability was assessed based on the characteristics of the development cohort and external validation cohorts. Risk of bias and concern regarding applicability were assessed based on the 'Prediction model study Risk Of Bias Assessment Tool' (PROBAST). RESULTS: Twelve studies fulfilled the inclusion criteria and covered 17 scoring systems across 5.130 patients. Several scores suffer from suboptimal development and validation. The SORG Nomogram, developed in a large surgical cohort, showed good discrimination on 3-month and 1-year survival, good calibration and was superior in direct comparison with low risk of bias and low concern regarding applicability. Machine learning algorithms are promising as they perform equally well in direct comparison. Tokuhashi, Tomita and other traditional risk scores showed suboptimal performance. CONCLUSION: The SORG Nomogram and machine learning algorithms outline superior performance in survival prediction for surgery in spinal metastases. Further improvement by comparative validation in large multicenter, prospective cohorts can still be obtained. Given the heterogeneity of spinal metastases, superior methodology of development and validation is key in improving future machine learning systems.

Cerebrovascular Autoregulation Monitoring in the Management of Adult Severe Traumatic Brain Injury: A Delphi Consensus of Clinicians.

BACKGROUND: Several methods have been proposed to measure cerebrovascular autoregulation (CA) in traumatic brain injury (TBI), but the lack of a gold standard and the absence of prospective clinical data on risks, impact on care and outcomes of implementation of CA-guided management lead to uncertainty. AIM: To formulate statements using a Delphi consensus approach employing a group of expert clinicians, that reflect current knowledge of CA, aspects that can be implemented in TBI management and CA research priorities. METHODS: A group of 25 international academic experts with clinical expertise in the management of adult severe TBI patients participated in this consensus process. Seventy-seven statements and multiple-choice questions were submitted to the group in two online surveys, followed by a face-to-face meeting and a third online survey. Participants received feedback on average scores and the rationale for resubmission or rephrasing of statements. Consensus on a statement was defined as agreement of more than 75% of participants. RESULTS: Consensus amongst participants was achieved on the importance of CA status in adult severe TBI pathophysiology, the dynamic non-binary nature of CA impairment, its association with outcome and the inadvisability of employing universal and absolute cerebral perfusion pressure targets. Consensus could not be reached on the accuracy, reliability and validation of any current CA assessment method. There was also no consensus on how to implement CA information in clinical management protocols, reflecting insufficient clinical evidence. CONCLUSION: The Delphi process resulted in 25 consensus statements addressing the pathophysiology of impaired CA, and its impact on cerebral perfusion pressure targets and outcome. A research agenda was proposed emphasizing the need for better validated CA assessment methods as well as the focused investigation of the application of CA-guided management in clinical care using prospective safety, feasibility and efficacy studies.

Estimated and underreported parameters in report based vehicle-bicycle accident reconstructions have a significant influence.

This study aimed to reconstruct four real life vehicle-bicycle collisions and evaluates the reconstruction parameters that affect the outcome of head injuries in report based accident reconstructions. A computational model of a car was developed in the multibody software MADYMO (MAthematical DYnamic MOdeling) and was used together with a validated bicycle model and the MADYMO 50 percentile pedestrian model. The accidents were reconstructed through an optimal fit method, based on kinematic and medical information. After the reconstruction, a parametric study on cyclists' movement and accident conditions was performed on the different cases. The velocity of the car and the angle of impact were found to significantly affect the accident outcome. This was demonstrated in terms of head injury criteria such as the Head Injury Criterion (HIC), the peak linear and peak angular velocity and acceleration. It was shown that the severity of the injury increases exponentially with increasing collision velocities. Additionally, the bicycle's parameters; crank rotation, handlebar angle and seat position revealed a large heterogeneity in the results. The maximum alteration between the lowest and highest HIC-value found for a complete crank rotation was a 416 % difference. For a handlebar rotation up to 100° or seat height alteration of maximum 34 cm, this value was respectively 169 % and 294 %. These high percentages of change indicate the need for cycling phase information for case-specific vehicle-bicycle accident reconstructions.

Autoregulation assessment by direct visualisation of pial arterial blood flow in the piglet brain.

Impairment of cerebrovascular autoregulation (CAR) is common after brain injury, although the pathophysiology remains elusive. The mechanisms of vascular dysregulation, their impact on brain function, and potential therapeutic implications are still incompletely understood. Clinical assessment of CAR remains challenging. Observational studies suggest that CAR impairment is associated with worse outcomes, and that optimization of cerebral blood flow (CBF) by individual arterial blood pressure (ABP) targets could potentially improve outcome. We present a porcine closed cranial window model that measures the hemodynamic response of pial arterioles, the main site of CBF control, based on changes in their diameter and red blood cell velocity. This quantitative direct CAR assessment is compared to laser Doppler flow (LDF). CAR breakpoints are determined by segmented regression analysis and validated using LDF and brain tissue oxygen pressure. Using a standardized cortical impact, CAR impairment in traumatic brain injury can be studied using our method of combining pial arteriolar diameter and RBC velocity to quantify RBC flux in a large animal model. The model has numerous potential applications to investigate CAR physiology and pathophysiology of CAR impairment after brain injury, the impact of therapeutic interventions, drugs, and other confounders, or to develop personalized ABP management strategies.

Serum neurofilament light chain levels as a marker of upper motor neuron degeneration in patients with Amyotrophic Lateral Sclerosis.

AIMS: Amyotrophic lateral sclerosis (ALS) is the most common motor neuron degeneration disease with a diagnostic delay of about 1 year after symptoms onset. In ALS, blood neurofilament light chain (NfL) levels are elevated, but it is not entirely clear what drives this increase and what the diagnostic performance of serum NfL is in terms of predictive values and likelihood ratios. The aims of this study were to further explore the prognostic and diagnostic performances of serum NfL to discriminate between patients with ALS and ALS mimics, and to investigate the relationship between serum NfL with motor neuron degeneration. METHODS: The diagnostic performances of serum NfL were based on a cohort of 149 serum samples of patients with ALS, 19 serum samples of patients with a disease mimicking ALS and 82 serum samples of disease control patients. The serum NfL levels were correlated with the number of regions (thoracic, bulbar, upper limb and lower limb) displaying upper and/or lower motor neuron degeneration. The prognostic performances of serum NfL were investigated based on a Cox regression analysis. RESULTS: The associated predictive values and likelihood ratio to discriminate patients with ALS and ALS mimics were established. Serum NfL was associated with motor neuron degeneration driven by upper motor neuron (UMN) degeneration and was independently associated with survival in patients with ALS. CONCLUSIONS: Altogether, these findings suggest that elevated serum NfL levels in ALS are driven by UMN degeneration and the disease progression rate and are independently associated with survival at time of diagnosis.

KidsBrainIT: A New Multi-centre, Multi-disciplinary, Multi-national Paediatric Brain Monitoring Collaboration.

OBJECTIVES: Validated optimal cerebral perfusion pressure (CPP) treatment thresholds in children do not exist. To improve the intensive care unit (ICU) management of the paediatric traumatic brain injury (TBI) population, we are forming a new paediatric multi-centre collaboration to recruit standardised ICU data for running and reporting upon models for assessing autoregulation and optimal CCP (CPPopt). MATERIALS AND METHODS: We are adapting the adult BrainIT group's approach to develop a new Paediatric Brain Monitoring and Information Technology Group (KidsBrainIT), which will include a repository to store prospectively collected high-resolution physiological, clinical, and outcome data. In the first phase of this project there are 7 UK Paediatric Intensive Care Units, 1 Spanish, 1 Belgium, and 1 Romanian Centre interested in participating. In subsequent phases, we plan to open recruitment to other centres both within Europe, US and abroad. We are collaborating with the Leuven Group and plan to use their LAx (low-frequency autoregulation index), DATACAR (dynamic adaptive target of active cerebral autoregulation), CPPopt and visualisation methodologies. We also plan to use the continuous diffuse optical monitoring and tomography technology developed in Barcelona as an acute surrogate end-point for optimising brain perfusion. This technology allows non-invasive continuous monitoring of deep tissue perfusion and oxygenation in adults but its clinical application in infants and children with TBI has not been studied previously. RESULTS: We report on the current status of setting up this new collaboration and also on pilot analyses in two centres which are the basis of our rationale for the need for a prospective validation study of CPPopt in children. Specifically, we demonstrated that CPPopt varied with time for each patient during their paediatric intensive care unit (PICU) stay, and the median overall CPPopt levels for children aged 2-6 years, 7-11 years and 12-16 years were 68.83, 68.09, and 72.17 mmHg respectively. Among survivors and patients with favourable outcome (GOS 4 and 5), there were significantly higher proportions with CPP monitoring time within CPPopt (p = 0.04 and p = 0.01 respectively). CONCLUSIONS: There is a need and an interest in forming a multi-centre PICU collaboration for acquiring data and performing analyses for determining validated CPPopt thresholds in the paediatric TBI population. KidsBrainIT is being formed to meet that need.

Cross-cultural adaptation and validation of the Dutch version of the core outcome measures index for low back pain.

PURPOSE: The core outcome measures index (COMI) is a validated multidimensional instrument for assessing patient-reported outcome in patients with back problems. The aim of the present study is to translate the COMI into Dutch and validate it for use in native Dutch speakers with low back pain. METHODS: The COMI was translated into Dutch following established guidelines and avoiding region-specific terminology. A total of 89 Dutch-speaking patients with low back pain were recruited from 8 centers, located in the Dutch-speaking part of Belgium. Patients completed a questionnaire booklet including the validated Dutch version of the Roland Morris disability questionnaire, EQ-5D, the WHOQoL-Bref, the Numeric Rating Scale (NRS) for pain, and the Dutch translation of the COMI. Two weeks later, patients completed the Dutch COMI translation again, with a transition scale assessing changes in their condition. RESULTS: The patterns of correlations between the individual COMI items and the validated reference questionnaires were comparable to those reported for other validated language versions of the COMI. The intraclass correlation for the COMI summary score was 0.90 (95% CI 0.84-0.94). It was 0.75 and 0.70 for the back and leg pain score, respectively. The minimum detectable change for the COMI summary score was 1.74. No significant differences were observed between repeated scores of individual COMI items or for the summary score. CONCLUSION: The reproducibility of the Dutch translation of the COMI is comparable to that of other validated spine outcome measures. The COMI items correlate well with the established item-specific scores. The Dutch translation of the COMI, validated by this work, is a reliable and valuable tool for spine centers treating Dutch-speaking patients and can be used in registries and outcome studies.

Surgery for metastatic spine tumors in the elderly. Advanced age is not a contraindication to surgery!

BACKGROUND: With recent advances in oncologic treatments, there has been an increase in patient survival rates and concurrently an increase in the number of incidence of symptomatic spinal metastases. Because elderly patients are a substantial part of the oncology population, their types of treatment as well as the possible impact their treatment will have on healthcare resources need to be further examined. PURPOSE: We studied whether age has a significant influence on quality of life and survival in surgical interventions for spinal metastases. STUDY DESIGN: We used data from a multicenter prospective study by the Global Spine Tumor Study Group (GSTSG). This GSTSG study involved 1,266 patients who were admitted for surgical treatments of symptomatic spinal metastases at 22 spinal centers from different countries and followed up for 2 years after surgery. PATIENT SAMPLE: There were 1,266 patients recruited between March 2001 and October 2014. OUTCOME MEASURES: Patient demographics were collected along with outcome measures, including European Quality of Life-5 Dimensions (EQ-5D), neurologic functions, complications, and survival rates. METHODS: We realized a multicenter prospective study of 1,266 patients admitted for surgical treatment of symptomatic spinal metastases. They were divided and studied into three different age groups: <70, 70-80, and >80 years. RESULTS: Despite a lack of statistical difference in American Society of Anesthesiologists (ASA) score, Frankel neurologic score, or Karnofsky functional score at presentation, patients >80 years were more likely to undergo emergency surgery and palliative procedures compared with younger patients. Postoperative complications were more common in the oldest age group (33.3% in the >80, 23.9% in the 70-80, and 17.9% for patients <70 years, p=.004). EQ-5D improved in all groups, but survival expectancy was significantly longer in patients <70 years old (p=.02). Furthermore, neurologic recovery after surgery was lower in patients >80 years old. CONCLUSIONS: Surgeons should not be biased against operating elderly patients. Although survival rates and neurologic improvements in the elderly patients are lower than for younger patients, operating the elderly is compounded by the fact that they undergo more emergency and palliative procedures, despite good ASA scores and functional status. Age in itself should not be a determinant of whether to operate or not, and operations should not be avoided in the elderly when indicated.

Multilevel cervical reconstruction with no remaining hardware: the potential of a vascularised fibular strut graft.

We present a case of a cervical chordoma extending from C3 to C6, treated by surgical resection followed by reconstruction in two stages with a vascularised fibular strut graft anteriorly and fibular allografts posteriorly. The goal was to achieve a stable reconstruction without any metal hardware to allow for increased accuracy of the adjuvant radiotherapy. The advantages of a vascularised fibula flap to realise this are discussed.

Traumatic brain injury in the elderly: a significant phenomenon.

INTRODUCTION: Traumatic brain injury (TBI) in the elderly is becoming an increasingly frequent phenomenon. Studies have mainly analyzed the influence of age as a continuous variable and have not specifically looked at geriatric patients as a group. The aim of this study is to map the magnitude and characteristics of geriatric TBI and to identify factors contributing to their poorer outcome. MATERIAL AND METHODS: Based on the ICD-9 register of the University Hospitals Leuven demographic and clinical variables of TBI were analyzed (2002-2008). The influence of older age on physiological variables was assessed using the Brain-IT database. RESULTS: The elderly (aged ≥65 years) accounted for 38.2% of non-concussion TBI and 32.6% of ICU admissions, representing the largest age group. The elderly had a significantly lower ICP (median 10.06 mmHg versus median 14.52 mmHg; p = 0.048), but no difference in their measure of autoregulation (daily mABP/ICP correlation coefficient) compared with 20-35 year-olds. TBI was caused by a fall in 78.9% of elderly patients and 42.3% suffered a mass lesion. 72.1% had cardiovascular comorbidity. Complications did not differ from their younger counterparts. DISCUSSION: Geriatric TBI is a significant phenomenon. Poorer outcomes are not yet sufficiently explained by physiological monitoring data, but reduced vascular versatility is likely to contribute. More research is needed in order to develop specific management protocols.

Oesophageal perforation after anterior cervical surgery: management in four patients.

BACKGROUND: Oesophageal perforation related to anterior cervical surgery is an uncommon but well recognised and potentially life-threatening complication with an incidence of 0-3.4%. Our experience with this complication and a review of the literature are presented. METHOD: We retrospectively reviewed our clinical experience over 10 years and found four patients in whom an oesophageal perforation was recognised after anterior surgery for cervical spine trauma. In three patients the perforation was noticed in the early post-operative period and the other had a delayed presentation. In all patients, the hardware was removed, long-term intravenous antibiotics were administered and parenteral nutrition was instituted. In two patients a primary suture of the perforation was performed and in one of these an additional sternocleidomastoid myoplasty was carried out as well. One patient had conservative treatment and one died before closure of the perforation could be performed. FINDINGS: The two patients, in whom surgical repair of the perforation was performed, recovered well with residual neurological deficits as expected due to the cervical trauma. In the patient in whom conservative treatment was instituted, healing of the perforation occurred. One patient died due to systemic complications, indirectly related to the perforation. CONCLUSIONS: Although not very frequent and sometimes difficult to diagnose, oesophageal perforations after anterior cervical surgery constitute a potentially life-threatening complication. Diagnosis is made by imaging or endoscopic studies, but clinical suspicion is most important. Basic treatment consists of surgery with removal of hardware, drainage of abscesses, primary closure of the perforation if possible, parenteral nutrition and antibiotic therapy. Residual instability should be recognised in time and may be anticipated in patients in whom there has been little time for solid bony fusion. Successful management depends on early diagnosis and immediate institution of treatment.

Lateral safety limits during a facial transclival approach to the craniovertebral junction.

BACKGROUND: Transclival procedures are infrequently performed and carry some dangerous complications. The carotid arteries and the hypoglossal nerves are the most important structures at risk during the subperiostal lateral dissection of the posterior wall of the oropharynx. These lateral landmarks are, however, not easily identified during the surgical dissection. METHODS: We measured the location of the foramen lacerum externum and the hypoglossal canal with respect to the midline. These measurements were performed on 76 non-pathological CT-scanned skulls that were converted into accurate three-dimensional digital reconstruction models by the use of 3D image-processing. The same measurements were also taken on 15 non-pathological skulls by the use of a slide calliper. RESULTS: For the foramen lacerum, the distance to the midline was 11 mm +/- 1 mm SD (range 8-14 mm) on the right-hand side and 11 mm +/- 1 mm SD (range 9-14 mm) on the left-hand side. For the right hypoglossal canal this distance was 17 mm +/- 1 mm SD (range 15-22 mm) and for the left hypoglossal canal 17 mm +/- 1 mm SD (range 14-20 mm). There was no significant difference between left and right. CONCLUSION: The posterior wall of the oropharynx can be safely dissected for at least 8 mm from the midline at the level of the foramen lacerum and for at least 14 mm towards the hypoglossal canal from the midline at the level of the anterior border of the foramen magnum. In addition, this technique with the help of 3D software, can be very useful in the pre-operative setting when performing complex skull base procedures.

A new test set-up for skull fracture characterisation.

Skull fracture is a frequently observed type of severe head injury. Historically, a variety of impact test set-ups and techniques have been used for investigating skull fracture. The most frequently used are the free-fall technique, the guided fall or drop tower set-up and the piston-driven impactor set-up. This document proposes a new type of set-up for cadaver head impact testing which combines the strengths of the most frequently used techniques and devices. The set-up consists of two pendulums, which allow for a 1 degree of freedom rotational motion. The first pendulum is the impactor and is used to strike the blow. The head is attached to the second pendulum using a polyester resin. Local skull deformation and impact force are measured with a sample frequency of 65 kHz. From these data, absorbed energy until skull fracture is calculated. A set-up evaluation consisting of 14 frontal skull and head impact tests shows an accurate measurement of both force and local skull deformation until fracture of the skull. Simplified mechanical models are used to analyse the different impacting techniques from literature as well as the new proposed set-up. It is concluded that the proposed test set-up is able to accurately calculate the energy absorbed by the skull until fracture with an uncertainty interval of 10%. Second, it is concluded that skull fracture caused by blunt impact occurs before any significant motion of the head. The two-pendulum set-up is the first head impact device to allow a well-controlled measurement environment without altering the skull stress distribution.

A clinical comparison of non-traumatic acute subdural haematomas either related to coagulopathy or of arterial origin without coagulopathy.

BACKGROUND: Non-traumatic acute subdural haematomas enable study of the morbidity and mortality due to the haematoma without the effect of trauma. Whereas it is known that coagulation disorders worsen the outcome of spontaneous intracerebral haematomas, this has not been studied in non-traumatic acute subdural haematomas. METHODS: In a series of 13 non-traumatic acute subdural haematomas admitted to our department between January 1995 and March 2002, we had 9 coagulopathy associated haematomas and 3 haematomas corresponding to the syndrome of 'spontaneous acute subdural haematoma of arterial origin'. Both groups were compared. FINDINGS: Age and gender distribution were comparable. The bleeding source was a cortical artery in 2 of the 2 non-coagulopathy related haematomas operated on, but also in 2 of the 4 coagulopathy associated haematomas that underwent surgery. The average haematoma thickness was higher in the coagulopathy related haematomas. The mean Glasgow Coma Score on admission was 7.7 and the mortality rate was 55.6% in the coagulopathy related group. In the non-coagulopathy related haematomas the mean Glasgow Coma Score was 12.0 and the mortality rate 33.3%. The latter mortality rate corresponds well to that of a historical group of 'spontaneous acute subdural haematomas of arterial origin' collected from the literature. INTERPRETATION: The outcome was worse in the non-traumatic acute subdural haematomas that were associated with a coagulation deficiency. While in all non-traumatic acute subdural haematomas the interval to surgery should be minimized, early recognition and urgent correction of coagulation deficiencies is certainly indicated.

Skull biomechanics: The energy absorbability of the human skull frontal bone during fracture under quasi-static loading.

Four human skulls were studied in order to determine the energy absorbed corresponding to a fracture due to quasi-static loading on the frontal bone. Using a dedicated experimental set up, the force-deformation characteristics of the specimens were recorded, calculating energy absorption. Mean values of 1975 +/- 703 N, 4.17 +/- 0.81 mm and 3.95 +/- 1.18 J were found for the peak force, maximum deformation and absorbed energy to fracture, respectively. Linear fractures were always seen in the frontal bone, with a similar pattern in three of the four skulls. These results can be used to develop bench-mark skull models to validate analytic models. (Journal of Applied Biomaterials & Biomechanics 2003; 1: 194-9).

Posterior fossa decompression in syringomyelia associated with a Chiari malformation: a retrospective analysis of 22 patients.

BACKGROUND: The technique of posterior fossa decompression, nowadays preferred by most neurosurgeons as the first surgical option in the treatment of hindbrain associated syringomyelia, was reintroduced in our center in 1989. We present a retrospective analysis of 22 patients with this pathology who underwent the procedure since then. METHODS: In all patients diagnosis was made by MRI of the craniospinal junction. The operation consisted of a suboccipital craniectomy, removal of the posterior arc of C1, laminectomy of C2 or C3 when necessary and a wide duraplasty at the end. The intradural manipulations (opening of the arachnoid membrane, coagulation or resection of the tonsils) were not uniform. Postoperatively short- and long-term clinical outcome and MRI findings were assessed. RESULTS: Sixteen out of 21 patients (76%) experienced an improvement in the early follow-up period. In the late follow-up period 13 out of 19 patients (68%) were improved, whereas five patients (26%) experienced a marked deterioration. There was no unequivocal effect on all symptoms and signs. Postoperative MR images showed a favorable result in 16 out of 20 patients (80%), consisting of syrinx collapse or reduction of the syrinx diameter. CONCLUSIONS: We conclude that decompression of the posterior fossa is a safe procedure with a considerable chance of clinical improvement. Although total syrinx collapse is not as frequently seen as in syrinx shunting procedures, the clinical outcome may be better. Moreover, there seemed to be no unequivocal correlation between clinical outcome and postoperative syrinx size in the present study.

Mesenteric venous thrombosis. Diagnostic and therapeutic approach.

Mesenteric venous thrombosis is an infrequent but distinct form of intestinal ischaemia. We report a case of acute mesenteric venous thrombosis diagnosed by computed tomography. Laparoscopy permitted to establish the extent of the ischaemia. Initially, high doses streptokinase were administered for 6 hours, followed by heparinotherapy for 10 hours, with the aim to reduce the length of bowel to be resected. One day later, intestinal resection was carried out, followed by postoperative anticoagulation.