|
 |
|
ORIGINAL ARTICLE |
|
|
|
| Year : 2020 | Volume
: 15
| Issue : 4 | Page : 409-415 |
| |
Clinical nomogram predicting intracranial injury in pediatric traumatic brain injury
Thara Tunthanathip1, Jarunee Duangsuwan2, Niwan Wattanakitrungroj2, Sasiporn Tongman3, Nakornchai Phuenpathom1
1 Division of Neurosurgery, Department of Surgery, Faculty of Medicine, Prince of Songkla University, Hat Yai, Thailand
2 Department of Computer Science, Faculty of Science, Prince of Songkla University, Hat Yai, Thailand
3 Department of Biotechnology, Faculty of Science and Technology, Thammasat University (Rangsit Campus), Khlong Luang, Thailand
| Date of Submission | 16-Jan-2020 |
| Date of Decision | 12-Mar-2020 |
| Date of Acceptance | 28-Mar-2020 |
| Date of Web Publication | 19-Jan-2021 |
Correspondence Address:
Dr. Thara Tunthanathip
Division of Neurosurgery, Department of Surgery, Faculty of Medicine Prince of Songkla University, Hat Yai, Songkhla 90110.
Thailand
 Source of Support: None, Conflict of Interest: None  | Check |
DOI: 10.4103/JPN.JPN_11_20
 Abstract | | |
Background: There are differences in injured mechanisms among pediatric traumatic brain injury (TBI) in developing countries. This study aimed to develop and validate clinical nomogram for predicting intracranial injury in pediatric TBI that will be implicated in balancing the unnecessary investigation in the general practice. Materials and Methods: The retrospective study was conducted in all patients who were younger than 15 years old and underwent computed tomography (CT) of the brain after TBI in southern Thailand. Injured mechanisms and clinical characteristics were identified and analyzed with binary logistic regression for predicting intracranial injury. Using random sampling without replacement, the total data was split into nomogram developing dataset (80%) and testing dataset (20%). Therefore, a nomogram was constructed and applied via the web-based application from the developing dataset. Using testing dataset, validation as binary classifiers was performed by various probabilities levels. Results: A total of 900 victims were enrolled. The mean age was 87.2 (standard deviation [SD] 57.4) months, and 65.3% of all patients injured were from road traffic accidents. The rate of positive findings in CT of the brain was 32.8%. A nomogram was developed from the significant variables, including age groups, road traffic accidents, loss of consciousness, scalp hematoma/laceration, motor weakness, signs of basilar skull fraction, low Glasgow Coma Scale score, and pupillary light reflex.Therefore, a nomogram was developed from 80% of data and was validated from 20% of data. The accuracy, sensitivity, specificity, positive, and negative predictive values of the nomogram were 0.83, 0.42, 1.00, 1.00, and 0.81 at a cutoff value of 0.5 probability. Conclusion: This study provides a clinical nomogram that will be applied to making decisions in general practice as a diagnostic tool from high specificity.
Keywords: Nomogram, pediatric traumatic brain injury, road traffic accident, traumatic brain injury
How to cite this article:
Tunthanathip T, Duangsuwan J, Wattanakitrungroj N, Tongman S, Phuenpathom N. Clinical nomogram predicting intracranial injury in pediatric traumatic brain injury. J Pediatr Neurosci 2020;15:409-15 |
| Introduction | |  |
From the global status report on road safety of the World Health Organization (WHO), Thailand is one of the high traffic-related death countries in the world. The common road user types in Thailand are motorcycles and vehicles, which lead to road accidents. The reported fatal incidence per 100,000 population of Thailand was 32.7 in 2018.[1],[2],[3] Children, who are the common group injured from road accidents in Thailand, suffer from neurological sequelae such as physical disability, cognitive dysfunction, and mortality.[4]
The computer tomography (CT) of the brain is presently the standard investigation for the evaluation of intracranial injury. However, the balance between radiation exposure from investigation and missing diagnosis of intracranial injury has been discussed for optimizing the investigation. To date, several clinical decision criteria have been derived in clinical practice; Canadian Assessment of Tomography for Childhood Head Injury (CATCH), Children’s Head Injury Algorithm for the Prediction of Important Clinical Events (CHALICE), and Paediatric Emergency Care Applied Research Network (PECARN).[5] However, the prior studies are based on pediatric traumatic brain injury (TBI) injured from falling, whereas the majority of patients in Thailand are injured from traffic accidents.[5]
Nomogram is a two-dimensional (2D) calculator designed from a mathematical function that has been applied to predicting various medical problems.[6],[7],[8],[9] Therefore, this study aimed to develop and validate a clinical nomogram for predicting intracranial injury in pediatric TBI that will be implicated in balancing the unnecessary investigation in the general practice.
| Materials and Methods | | |
Study designs and study population
The authors conducted a retrospective cohort study in all patients who were younger than 15 years old and underwent a CT of the brain after TBI at Songklanagarind Hospital, Prince of Songkla University, between January 2009 and December 2018. The study was approved by the institutional committees on research ethics (REC. 62-065-10-1).
Data collection
The following demographic data, various clinical findings, mechanisms of injury, laboratory, and treatment, were obtained from the electronic medical database for analysis. The initial pediatric Glasgow Coma Scale (GCS) score on arrival was divided according to severity.[10] In detail, road traffic injury included patients who were either the driver or passenger on motorbikes, vehicles, or pedestrians or who were crashed by any vehicles.
All patients reporting with results of CT of the brain were performed as binary classifiers. Positive findings defined at least one type of intracranial injury was observed; skull fracture, basilar skull fracture, cerebral contusion, subdural hematoma, epidural hematoma, subarachnoid hemorrhage, and intraventricular hemorrhage. In addition, midline shift and patency of basal cistern were evaluated from the medical database.
Statistical analysis
The proportion means with standard deviation (SD) were calculated for descriptive purposes from the full dataset. Therefore, the total data were split into nomogram developing dataset (80%) and performance testing dataset (20%) by random sampling without replacement.
Univariate and multivariable analyses with binary logistic regression were performed from developing dataset to identify the factors associated with the intracranial injury, which were divided into two groups (positive and negative groups). A value of P < 0.05 was considered statistically significant. All factors included in the final models were a test for interactions with each factor. For the collinearity evaluation of each factor, the tolerance values and the variance inflation factor for each factor were >0.5 and <10, respectively. Therefore, the goodness-of-fit test was assessed using the Hosmer–Lemeshow test, whereas the discrimination of the model was evaluated by the concordance index. The statistical analysis was performed using the R version 3.4.0 software (R Foundation, Vienna, Austria).
Nomogram development and validation
The final model from the multivariable analysis was created to be the nomogram predicting intracranial injury. The internal model validity was tested by the bootstrap method with 1000 iterations. The “rms” package was used to develop the nomogram.[11] The 2D nomogram was converted to be digital nomogram by the shiny package version 1.4.0.[12],[13]
Therefore, the predictive abilities of the nomogram were evaluated using the testing dataset. The receiver-operating characteristic (ROC) curve and the largest area under the receiver-operating characteristic (AUC) were performed the optimal cutoff point in each nomogram. Furthermore, the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were determined in each nomogram.[14]
| Results | | |
Demographic characteristics
One thousand, one hundred and ninety-eight pediatric patients were diagnosed with TBI between January 2009 and December 2018. In detail, patients who diagnosed between 2009 and 2015 were the study population of Tunthanathip and Phuenpathom,[4] and TBI patients diagnosed between 2016 and 2018 were added in this study. Of these, two hundred and ninety-eight patients were excluded because they did not go to neuroimaging investigations, leaving the study population at 900 patients. The baseline characteristics of the study population were described in [Table 1]. Over two-thirds of the pediatric patients with TBI were males with a mean age of 87.2 (SD 57.4) months. The mechanism of injury, over two-thirds (65.3%) of patients were injured from road traffic accidents, 60.7% of traffic injured patients were a passenger, whereas 22.4% were motorcycle drivers. The mean age of the drivers was 160.3 months (SD 21, range 100–192), and patients older than 60 months had a higher prevalence of traffic accidents than patients 60 months or younger (chi-square test, P < 0.001). Common injuries were scalp wound/hematoma, loss of consciousness, and amnesia. The seizure was observed as the first presentation in 4.9% of patients. The majority of injured children were mild TBI, whereas moderate and severe TBI were 8.7% and 9.1%, respectively. Abnormality of pupillary light reflex was observed in 6% of all patients, and hypotension was a significantly higher risk to abnormal pupillary light reflex in both eyes (chi-square test, P < 0.001)
As a result, the positive rate of investigations was 32.8% of all imagings. In detail, the most common intracranial injury was a skull fracture, whereas subdural hematoma was observed in 13.3% of all investigations.
Factors associated with intracranial injury
The development of a nomogram was constructed from 720 patients, and the tool was developed from preoperative factors associated with the intracranial injury. In univariate analysis, the independent predictors included age older than 60 months (odds ratio [OR] 1.81, 95% confidence interval (CI) 1.29–2.54), road traffic accident (OR 2.12, 95%CI 1.49–3.02), loss of consciousness (OR 1.60, 95%CI 1.16–2.20), motor weakness (OR 10.10, 95%CI 3.38–30.22), scalp wound (OR 1.68, 95%CI 1.21–2.34), bleeding per nose/ear (OR 22.77, 95%CI 5.27–98.28), hypotension (OR 3.31, 95%CI 1.56–7.00), GCS 9–12 (OR 3.35, 95%CI 1.92–5.85), GCS 3–8 (OR 18.75, 95%CI 9.35–37.57), fixed pupillary response in one eye (OR 12.27, 95%CI 3.51–42.85), and fixed pupillary response in both eyes (OR 17.99, 95%CI 5.3–60.81) as shown in [Table 2]. | Table 2: Binary logistic regression analysis for intracranial injury from developing dataset
Click here to view |
Therefore, the final model comprised age older than 60 months (OR 1.60, 95%CI 1.07–2.41), road traffic accident (OR 1.34, 95%CI 0.88-20.50), loss of consciousness (OR 1.47, 95%CI 1.001-2.17), motor weakness (OR 5.76, 95%CI 1.64-20.17), scalp wound (OR 2.79, 95%CI 1.84-4.22), bleeding per nose/ear (OR 28.49, 95%CI 6.03-134.55), GCS 9–12 (OR 3.47, 95%CI 1.88-6.40), GCS 3–8 (OR 11.26, 95%CI 4.94-25.65), fixed pupillary response in one eye (OR 10.12, 95%CI 2.56–39.9), fixed pupillary response in both eyes (OR 3.18, 95%CI 0.73–13.75). Although road traffic accident was not significant in multivariable analysis, this issue is one of the important mechanisms in several countries. Also, this factor was included in the final model.
The P value of the Hosmer–Lemeshow test for goodness of fit was 0.11, whereas the concordance index of the model discrimination was 0.788.
Nomogram development
Nomogram was constructed from multivariable analysis, as shown in [Figure 1]. Therefore, the bootstrapped calibration plot showed that the model was very close to the ideal as [Figure 2]. For general practice, the 2D nomogram was created to be the web-based application via https://thara.shinyapps.io/tsus4/. The result of prediction was reported as the probability of a positive finding of the CT ranging from 0 to 1 with 95%CI. The predictive abilities of the nomogram were evaluated according to 0.4, 0.5, and 0.6 of probabilities of positive results. | Figure 1: Clinical nomogram predicting the intracranial injury of pediatric traumatic brain injury. To use the nomogram, draw a straight line upward from the patient’s characteristics of age, injury mechanism, loss of consciousness (LOC), motor weakness, scalp injury, bleeding per nose/ear, Glasgow Coma Score (GCS), and pupillary light reflex to the upper points scale. Take the sum of the points received for each predictor and locate this sum on the total points scale. Draw a straight line down to the predictive value scale to find the patient’s probability of intracranial injury
Click here to view |
 | Figure 2: Bootstrapped calibration plot, which proves that concordance between the predicted probability and response is satisfactory
Click here to view |
In detail, the predictive ability of each cutoff value composed sensitivity, specificity, PPV, NPV, accuracy, the ROC curve, and the AUC as shown in [Table 3]. | Table 3: Abilities of nomogram predicting intracranial injury according to each probability level from testing dataset
Click here to view |
At a cutoff value of the 0.5 probability (predicted intracranial injury when predictive probability from equaled 0.5 or more), the nomogram had a sensitivity of 0.42, a specificity of 1.00, PPV of 1.00, NPV of 0.81 and accuracy of 0.83, and AUC of 0.71. As shown in [Figure 3], the cutoff value is the best abilities, whereas the nomogram’s ability at the cutoff value of 0.4 probability dropped in all aspects. | Figure 3: ROC curve and AUC of nomogram each cutoff value of the probability of the prediction
Click here to view |
| Discussion | | |
A fall from standing or from height was the common mechanism of injury in children between 17% and 68%, whereas traffic injury had been reported as <5%.[15],[16],[17],[18],[19] However, the high prevalence of traffic-related mortality has been observed in several developing counties.[1],[2] Osifo et al.[20] studied mortality in pediatric patients from road traffic accidents in Nigeria’s referral center found that 9% of all cases were the leading cause of fatal accidents. Similarly, the majority of injured mechanism in this study was road traffic accidents included drivers, passengers, and pedestrians, and these accidents significantly led to the intracranial injury. Because of the significant differences in dangerous mechanisms that were found in these counties, the clinical decision rule to investigate individually is developed to apply in general practice.[4]
Intracranial injury and associated factors
Clinical variables were essential to predicting intracranial injuries in children. Scalp hematoma/laceration, loss of consciousness, motor weakness, signs of basilar skull fraction, low GCS score, and pupillary light reflex were significantly associated with the intracranial injury. Noticeably, abnormal pupillary light reflex has a significantly higher risk of intracranial injury than normal pupillary response, especially unilateral fixed pupil. Because nonreactive pupils can also result from systematic disturbance such as hypoxia, hypotension than brainstem compression while a fixed pupil in one eye directly reflected a unilateral brainstem compression or ipsilateral oculomotor nerve compression as uncal herniation.[21],[22],[23],[24],[25]
The pupillary light reflex has the neural circuits in the intrinsic brainstem that the pathway conducted by optic tract fibers synapsing in the pretectal area. Therefore, the pretectal neurons transfer to the Edinger–Westphal complex, and the ciliary nerve serves pupillary constrictor reactions.[10] Ritter reported that pupillary dilatation and nonreactive pupils caused by decreasing brainstem blood flow as 30.5 ± 16.8mL/100g/min, whereas normal pupillary response had brainstem blood flow as 43.8 ± 18.7mL/100g/min.[21]
Nomogram validation
A nomogram is a simple tool of a scoring system, which was used to predict several medical conditions such as metabolic condition,[5] degenerative diseases,[6] and TBI.[8] However, the 2D nomogram has limitations about convenient usage. Therefore, we proposed a user-friendly digital nomogram for real-world practice. Also, the best predictive ability of the nomogram was addressed at 0.5 probability as the cutoff value. In detail, this cutoff value had the highest specificity, PPV, and AUC, which is an acceptable level.[12] For real-world applications, the high-specificity and high PPV tool are appropriate to be a diagnostic tool. Alternative methods such as various machine learning algorithms were reported that achieved reasonable performance for predicting non-big data.[25] Wong et al.[1] reported that K-nearest neighbor, logistic regression, and random forest algorithms could produce accurate disease predictability better than traditional regression analysis in prostate cancer. In addition, the study by Singal et al.[26] reported that the machine learning algorithm improves the accuracy of risk stratifying patients with cirrhosis. The improvement of the tool’s ability with the machine learning algorithm is a challenge that should be conducted in the future.[27],[28],[29]
The strength of this study includes a large number of pediatric TBI cases that injured from more severe mechanisms than the Western country as a road traffic accident. As the authors’ knowledge, this study is the first paper proposed the digital nomogram predicting intracerebral injuries of children after TBI. This tool was simplified and easy to use for bedside implications. However, the external validation of a web-based application should be estimated in the future.
Limitations of this study should be acknowledged that the retrospective design of our study is a limitation. As with any observations study, the possibility of confounders cannot be excluded. Therefore, a multivariable analysis was performed to tackle the limitation.[30] Besides, a low prevalence of some clinical variables, such as seizure, obscured the risk of intracranial injury.
| Conclusion | | |
Road traffic accidents carry out intracranial injury among children. This study provides a clinical nomogram that can be applied in general practice. Moreover, implication with the prospective study needs to conduct further studies, or a machine-learning application may be an alternative method to enhance predictive performance.
Ethical policy and institutional review board statement
All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards (REC. no. 62-065-10-1).
Acknowledgement
We would like to thank associate professor Paramee Thongsuksai for their advice about manuscript preparation.
Transparency declaration
The study population were TBI patients younger than 15 years old and underwent a CT of the brain between January 2009 and December 2018. In detail, patients who diagnosed between 2009 and 2015 were study population of Tunthanathip and Phuenpathom.[4] and TBI patients diagnosed between 2016 and 2018 were added in this study.
Financial support and sponsorship
This study was funded by Targeted Research Gants, Faculty of Medicine, Prince of Songkla University (Grant no. 62-065-10-1).
Conflicts of interest
There are no conflicts of interest.
| References | | |
| 1. | Wong NC, Lam C, Patterson L, Shayegan B Use of machine learning to predict early biochemical recurrence after robot-assisted prostatectomy. BJU Int 2019;123:51-7.  |
| 2. | World Health Organization. Global status report on road safety 2018. Available from: https://apps.who.int/iris/bitstream/handle/10665/277370/WHO-NMH-NVI-18.20-eng.pdf?ua=1. [Last accessed on 2019 May 29]. |
| 3. | World Health Organization. Death on the roads. Available from: https://extranet.who.int/roadsafety/death-on-the-roads/#country_or_area/THA. [Last accessed on 2019 May 29]. |
| 4. | Tunthanathip T, Phuenpathom N Impact of road traffic injury to pediatric traumatic brain injury in southern Thailand. J Neurosci Rural Pract 2017;8:601-8. |
| 5. | Easter JS, Bakes K, Dhaliwal J, Miller M, Caruso E, Haukoos JS Comparison of PECARN, CATCH, and CHALICE rules for children with minor head injury: a prospective cohort study. Ann Emerg Med 2014;64:145-52, 152.e1-5. |
| 6. | Devin CJ, Bydon M, Alvi MA, Kerezoudis P, Khan I, Sivaganesan A, et al. A predictive model and nomogram for predicting return to work at 3 months after cervical spine surgery: an analysis from the quality outcomes database. Neurosurg Focus 2018;45:E9. |
| 7. | Dogan HS, Altan M, Citamak B, Bozaci AC, Karabulut E, Tekgul S A new nomogram for prediction of outcome of pediatric shock-wave lithotripsy. J Pediatr Urol 2015;11:84.e1-6. |
| 8. | Tunthanathip T, Udomwitthayaphiban S Development and validation of a nomogram for predicting the mortality after penetrating traumatic brain injury. Bull Emerg Trauma 2019;7:347-54. |
| 9. | Liang W, Zhang L, Jiang G, Wang Q, Liu L, Liu D, et al. Development and validation of a nomogram for predicting survival in patients with resected non-small-cell lung cancer. J Clin Oncol 2015;33:861-9. |
| 10. | Kirkham FJ, Newton CR, Whitehouse W Paediatric coma scales. Dev Med Child Neurol 2008;50:267-74. |
| 11. | Harrell FE Jr. Package “rms”. Available from: https://cran.r-project.org/web/packages/rms/rms.pdf. [Last accessed on 2019 Jan 21]. |
| 12. | Chang W, Cheng J, Allaire JJ, Xie Y, McPherson J shiny: Web Application Framework for R. Available from: https://cran.r-project.org/web/packages/shiny/index.html. [Last accessed on 2020 Feb 21]. |
| 13. | Jalali A, Alvarez-Iglesias A, Roshan D, Newell J Package ‘DynNom’. Available from: https://cran.r-project.org/web/packages/DynNom/DynNom.pdf. [Last accessed on 2020 Feb 21]. |
| 14. | Swets JA ROC analysis applied to the evaluation of medical imaging techniques. Invest Radiol 1979;14:109-21. |
| 15. | Kraus JF, Rock A, Hemyari P Brain injuries among infants, children, adolescents, and young adults. Am J Dis Child 1990;144:684-91. |
| 16. | Kuppermann N, Holmes JF, Dayan PS, Hoyle JD Jr, Atabaki SM, Holubkov R, et al; Pediatric Emergency Care Applied Research Network (PECARN). Identification of children at very low risk of clinically-important brain injuries after head trauma: a prospective cohort study. Lancet 2009;374:1160-70. |
| 17. | Osmond MH, Klassen TP, Wells GA, Correll R, Jarvis A, Joubert G, et al; Pediatric Emergency Research Canada (PERC) Head Injury Study Group. CATCH: a clinical decision rule for the use of computed tomography in children with minor head injury. CMAJ 2010;182:341-8. |
| 18. | Pontarelli EM, Jensen AR, Komlofske KM, Bliss DW Infant head injury in falls and nonaccidental trauma: Does injury pattern correlate with mechanism? Pediatr Emerg Care 2014;30:677-9. |
| 19. | Trefan L, Houston R, Pearson G, Edwards R, Hyde P, Maconochie I, et al. Epidemiology of children with head injury: a national overview. Arch Dis Child 2016;101:527-32. |
| 20. | Osifo OD, Osagie TO, Iribhogbe PE Pediatric road traffic accident deaths presenting to a Nigerian referral center. Prehosp Disaster Med 2012;27:136-41. |
| 21. | Chaudhuri K, Malham GM, Rosenfeld JV Survival of trauma patients with coma and bilateral fixed dilated pupils. Injury 2009;40:28-32. |
| 22. | Chen JW, Gombart ZJ, Rogers S, Gardiner SK, Cecil S, Bullock RM Pupillary reactivity as an early indicator of increased intracranial pressure: the introduction of the neurological pupil index. Surg Neurol Int 2011;2:82. |
| 23. | Ritter AM, Muizelaar JP, Barnes T, Choi S, Fatouros P, Ward J, et al. Brain stem blood flow, pupillary response, and outcome in patients with severe head injuries. Neurosurgery 1999;44:941-8. |
| 24. | Stevens RD, Shoykhet M, Cadena R Emergency neurological life support: intracranial hypertension and herniation. Neurocrit Care 2015;23:S76-82. |
| 25. | Panesar SS, D’Souza RN, Yeh FC, Fernandez-Miranda JC Machine learning versus logistic regression methods for 2-year mortality prognostication in a small, heterogeneous glioma database. World Neurosurg X 2019;2:100012. |
| 26. | Singal AG, Mukherjee A, Elmunzer BJ, Higgins PD, Lok AS, Zhu J, et al. Machine learning algorithms outperform conventional regression models in predicting development of hepatocellular carcinoma. Am J Gastroenterol 2013;108:1723-30. |
| 27. | Tunthanathip T, Phuenpathom N, Sae-Heng S, Oearsakul T, Sakarunchai I, Kaewborisutsakul A Traumatic cerebrovascular injury: clinical characteristics and illustrative cases. Neurosurg Focus 2019;47:E4. |
| 28. | Owlia M, Yu L, Deible C, Hughes MA, Jovin F, Bump GM Head CT scan overuse in frequently admitted medical patients. Am J Med 2014;127:406-10. |
| 29. | Parma C, Carney D, Grim R, Bell T, Shoff K, Ahuja V Unnecessary head computed tomography scans: a level 1 trauma teaching experience. Am Surg 2014;80:664-8. |
| 30. | Mann CJ Observational research methods. Research design II: cohort, cross sectional, and case-control studies. Emerg Med J 2003;20:54-60. |
[Figure 1], [Figure 2], [Figure 3]
[Table 1], [Table 2], [Table 3]
| This article has been cited by | | 1 |
Factors Associated with Recurrence in Chronic Subdural Hematoma following Surgery |
|
| Kritsada Buakate, Thara Tunthanathip | | Journal of Health and Allied Sciences NU. 2023; | | [Pubmed] | [DOI] | | | 2 |
Economic impact of a machine learning-based strategy for preparation of blood products in brain tumor surgery |
|
| Thara Tunthanathip, Sakchai Sae-heng, Thakul Oearsakul, Anukoon Kaewborisutsakul, Chin Taweesomboonyat, Venkatesh Shankar Madhugiri | | PLOS ONE. 2022; 17(7): e0270916 | | [Pubmed] | [DOI] | | | 3 |
Comparison of intracranial injury predictability between machine learning algorithms and the nomogram in pediatric traumatic brain injury |
|
| Thara Tunthanathip, Jarunee Duangsuwan, Niwan Wattanakitrungroj, Sasiporn Tongman, Nakornchai Phuenpathom | | Neurosurgical Focus. 2021; 51(5): E7 | | [Pubmed] | [DOI] | | | 4 |
Application of machine learning to predict the outcome of pediatric traumatic brain injury |
|
| Thara Tunthanathip, Thakul Oearsakul | | Chinese Journal of Traumatology. 2021; 24(6): 350 | | [Pubmed] | [DOI] | | | 5 |
Prognostic Impact of the Combination of MGMT Methylation and TERT Promoter Mutation in Glioblastoma |
|
| Thara Tunthanathip, Surasak Sangkhathat, Pimwara Tanvejsilp, Kanet Kanjanapradit | | Journal of Neurosciences in Rural Practice. 2021; 12(04): 694 | | [Pubmed] | [DOI] | |
|
|
|
|
|
|
|
|
