Development of data mining method for medical prediction in neurology
-
1
University Hospital Lenval, France
-
2
University of L'Aquila, Italy
In our clinical practice we often wonder what might be the preceding clinical signs in patients with complex neurological disorders and how can we prevent before even to treat them. Comorbidity represents one of major issues during our clinical experience (1).
For example neurological conditions as epilepsy are often associated with many other diseases (2). We lack knowledge of clinical appearance of epilepsy and its putative role on clinical course of patients with several additional disorders.
OBJECTIVES: Our aim is to create a method to identify factors that could predict in advance diseases and clinical features in early stages co-occurring with epilepsy in patients with neurodevelopmental disorders.
We created an electronic database with patient’s data, and then we performed a data analysis using statistical and machine learning methods.
It could have many advantages, including reducing the cost of care for a longer-lasting well-being. We are studying the implementation of a statistical and medical data mining method in order to reach this goal.
METHODS: This is a cross-sectional descriptive study. It was carried out in the University Pediatric Hospital Lenval of Nice. We examined data of 70 hospitalized subjects with neurodevelopmental disorders during 10years (2004-2014), mean age of 16.8 years (range, 12-18 years).
The inclusion criteria were as follows: (a) classification of pyramidal disorders by neurological evaluation according to the Surveillance of Cerebral Palsy in Europe proposal, (b) chronological age between 12 and 18 years, (c) at least 3 years follow up d) presence of severe intellectual impairment (defined by IQ<50).
The exclusion criteria were: progressive or spinal neuro-pathologies, developed gross motor functional capacities (defined by GMFM >70), and manual abilities (MACS >3).
Etiological causes were listed in primary and annexed. All patients underwent structured battery functional assessments as Terver score for motor functional assessment, Rancho Los Amigos Level of cognitive functioning Scale, Zausmer Evaluation of motor development, Manual Ability Classification System , Gross Motor Function Classification System, Bohannon and Smith Modified Ashworth Scale for Spasticity and Modified Tardieu Scale for spasticity and clinical assessment. We also noted the presence of extrapyramidal disorders, and type of epilepsy (mild versus moderate and severe). Orthopedic assessment was performed with goniometric measurements, radiography ratings, and clinical setting.
We have created an electronic database, so-called “Bertoncelli Database” with a list of detailed information about each of the 70 participants. We transcribed their personal details and etiological information, diagnostic, functional and neurological data, whole body assessment of spastic hypertonia, and psychopathological advices.
Statistical analysis
Statistical analysis was carried on participant’s data which were divided into two steps.
For the first step we used contingency tables and Fisher test to identify risk factors, confidence interval and distribution frequency of two variables: epilepsy yes/no versus etiological data, functional and neurological diseases.
In ordinates we have noted the presence or absence of epilepsy, in the abscissa the presence or absence of functional disorders and / or neurological diseases.
For this purpose we used “OpenEpi” software, a web based epidemiologic calculator (3).
We’ve considered being statistically significant, the minimum number of cases required at the 95% confidence interval (p.0.5).
As second step we used logistic regression to create a regression model capable of predicting the probability of developing each disease (the dependent variable) based on the values of independent variables previously identified in the first step as risk factors.
For this purpose we used statistical open source software “R” (4,5) and its general linear model glm() function. The average prediction accuracy resulted to be 75%. We are confident to significantly improve this value in the future by optimizing the algorithm on a larger data-set of patients (110 in program).
RESULTS: It is well known that subjects with neurodevelopmental disorders have an high risk of having epilepsy associated with other diseases (6, 7).
More specifically we wave found that patients with controlled epilepsy compared with those without epilepsy have more chances to develop pyramidal disorders (P = 0.0263), Upper (P = 0.0498) and Lower (P = 0.013) extremities spasticity, ankle spastic hypertonia (P = 0.0110) and have a higher risk to not walk undependably (P = 0.0105).
Patients with non-tractable epilepsy, have more risk to develop scoliosis compared with those non-epileptic (P= 0.0079) and with controlled epilepsy (P = 0.0104).
About etiological predictors we have noted that only epileptic subjects with neurodevelopmental disorders due to post-natal causes have an higher risk to develop in future non-tractable epilepsy compared with those with pre-natal causes (P = 0.0052).
CONCLUSIONS: We have created a method that could predict some pathological manifestation in epileptic subjects with neurodevelopmental disorders based on statistical analysis of their clinical data. Type of epilepsy, diagnosed several years before specific dysfunctions, may represent a “predictor” of comorbidities.
Our next step is to apply this evidence-based method to predict clinical features on a larger number of participants (110), using logistic regression and other statistical or machine learning methods.
References
1. Meredith R, Golomb, MD, Chandan Saha, et al. The Association of Cerebral Palsy with Other Disability in Children with Perinatal Arterial Ischemic Stroke. Pediatr Neurol. 2007; 37: 245–249.
2. Hundozi-Hysenaj H, Boshnjaku-Dallku I. Epilepsy in children with cerebral palsy. J Pedia Neurol. 2008; 6:43–46.
3. Sullivan K, Andrew D., Minn Minn S. OpenEpi: a web-based epidemiologic and statistical calculator for public health. Public Health Reports. 2009; 124: 458- 471.
4. Team R., Core R. A Language and Environment for Statistical Computing. R Foundation for Statistical Computing. Vienna, Austria; 2012. ISBN 3-900051- 07-0, 2012. 20
5. Hilbe J. Logistic Regression Using R. Book Medical Publishers;2009.
6. Bruck I, Antoniuk SA, Spessatto A, Bem RS, Hausberger R, Pacheco CG. Epilepsy in children with cerebral palsy. Arq Neuropsiquiatr.2001; 59:35–39.
7. Carlsson M, Hagberg G, Olsson I. Clinical and aetiological aspects of epilepsy in children with cerebral palsy. Dev Med Child Neurol.2009; 45:371– 376.
Keywords:
Epilepsy,
Neuropediatrics,
statistics,
Comorbidity,
prediction
Conference:
SAN2016 Meeting, Corfu, Greece, 6 Oct - 9 Oct, 2016.
Presentation Type:
Oral Presentation in SAN 2016 Conference
Topic:
Oral Presentations
Citation:
BERTONCELLI
C and
BERTONCELLI
D
(2016). Development of data mining method for medical prediction in neurology.
Conference Abstract:
SAN2016 Meeting.
doi: 10.3389/conf.fnhum.2016.220.00029
Copyright:
The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers.
They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.
The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.
Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.
For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.
Received:
29 Jul 2016;
Published Online:
30 Jul 2016.
*
Correspondence:
Prof. CARLO BERTONCELLI, University Hospital Lenval, NICE, France, bertoncelli@unice.fr