^{1}

^{*}

^{2}

^{3}

^{4}

^{2}

^{5}

Many preterm infants suffer from neural disorders caused by early birth complications. The detection of children with neurological risk is an important challenge. The electroencephalogram is an important technique for establishing long-term neurological prognosis. Within this scope, the goal of this study is to propose an automatic detection of abnormal preterm babies’ electroencephalograms (EEG). A corpus of 316 neonatal EEG recordings of 100 infants born after less than 35 weeks of gestation were preprocessed and a time series of standard deviation was computed. This time series was thresholded to detect Inter Burst Intervals (IBI). Temporal features were extracted from bursts and IBI. Feature selection was carried out with classification in one step so as to select the best combination of features in terms of classification performance. Two classifiers were tested: Multiple Linear Regressions and Support Vector Machines (SVM). Performance was computed using cross validations. Methods were validated on a corpus of 100 infants with no serious brain damage. The Multiple Linear Regression method shows the best results with a sensitivity of 86.11% ± 10.01%, a specificity of 77.44% ± 7.62% and an AUC (Area under the ROC curves) of 0.82 ± 0.04. An accurate detection of abnormal EEG for preterm infants is feasible. This study is a first step towards an automatic analysis of the premature brain, making it possible to lighten the physician’s workload in the future.

About 15 million newborns are born prematurely every year in the world [

During the past four decades, several studies exploited preterm babies EEG to study neural disorders. Intensive studies focused on the neurological outcome of neonatal EEG [

aEEG is an accurate method for establishing long-term neurological prognosis with sensitivities and specificities comparable to cerebral ultrasound assessment. In [

Several studies tried to automatize bursts detection and seizures occurrences (uncontrolled electrical activity in the brain, producing physical convulsions, minor physical signs, thought disturbance, or a combination of those symptoms). For instance, authors of [

Finally we would like to quote a recent work we did on this problematic [

The method outlined in this paper works in four steps. First, a preprocessing stage: EEG was filtered, using a band-stop IIR filter and smoothed using a moving average window. Secondly, IBI were detected by thresholding standard deviation of preprocessed EEG. Thirdly, temporal features were extracted from IBI and bursts. Finally, feature selection was incorporated in the classification step so as to select relevant features that maximize classification performance. Two classifiers were tested: Support Vector Machines (SVM) and Multiple Linear Regressions with all combinations of features. Performance measures were evaluated using areas under the ROC curves (AUC, [

The paper is outlined as follows: Section 2 describes the collected database. Section 3 accounts for the method. While Section 4 describes the results, Section 5 provides a discussion. Finally a conclusion is drawn and some future works are suggested.

EEG signals from 100 preterm infants were collected in the Hospital of Angers, France, at the neonatal intensive care unit of the neuropediatric department. This monitoring was part of the usual clinical follow up of premature infants. All legal representatives of the babies gave informed consent for participation in research studies. EEG were recorded at sampling rate of 256 Hz. The recording system (Alliance from Nicolet Biomedical) was used with 8 to 11 adapted scalp electrodes according to the head size. Therefore, each EEG was composed of 11 channels. Electrodes were placed according to the international 10 to 20 system (

Thus, 416 neonatal EEG recordings lasting from 30 to 45 minutes were performed between January 1, 2003 and December 31, 2004. All 100 infants had less than 35 weeks of gestation. Each baby had between 1 to 7 recordings.

The 416 EEG were reviewed by a neuropediatrician expert and classified as normal, abnormal and doubtful. Thus by a careful visual analysis, EEG were considered normal if the background activity, in relation to the gestation age, was normal and no abnormal features on the EEG appeared. The abnormal EEG were those who showed excessive discontinuities with maximal IBI duration greater than 50% of the maximal value (in relation to the age of gestation), seizures or positive rolandic sharp waves of more than 2 per minute. From 416 EEG, 100 EEG recordings were considered as doubtful and were thus rejected. Finally, for the 316 kept EEG, the careful visual eye inspection led to 274 normal EEG (88.77%, 31.04 ± 2.13 weeks of gestation) and 42 abnormal EEG (11.23%,

30.01 ± 2.19 weeks of gestation). An example of abnormal EEG is illustrated in

Let s ( t ) denoting the EEG signal of N samples recorded in a given channel, in which abnormal EEG have to be detected. The EEG signal essentially contains background activity where bursts appear together with abnormal activities (IBI with discontinuity, seizures, rolandic sharp waves, etc.). The problem we address in this paper consists first in detecting the IBI and secondly classifying EEG into normal or abnormal. Automatic detection of abnormal EEG works in four steps summarized in

For each channel, raw EEG signal s ( t ) has been band-stop filtered at 50 Hz with a notch second order Butterworth IIR filter. Thus, we obtained a filtered signal s B P ( t ) where the power supply frequency of 50 Hz was removed. Then, s B P ( t ) has been smoothed by calculating the moving average over a window of width ω 1 :

s M A [ n ] = 1 ω 1 ∑ k = n − ω 1 / 2 n + ω 1 / 2 s B P [ k ] , n = 1 , ⋯ , N (1)

For detecting IBI, the standard deviation of signal s M A ( t ) has been computed and thresholded as in the work of [

ν 2 [ n ] = 1 ω 2 − 1 ∑ k = n ω 3 n ω 3 + ω 2 − 1 s M A 2 [ k ] − 1 ω 2 ( ω 2 − 1 ) ( ∑ k = n ω 3 n ω 3 + ω 2 − 1 s M A [ k ] ) 2 , n = 1 , ⋯ , N (2)

Successive standard deviation segments with values less than a threshold V T (in μV) and longer than 1 s have been detected and delimited by an onset and an offset boundary limit markers. Consecutive detections less than 0.5 s apart have been grouped together and considered as the same IBI. Finally, only IBI present across all 11 EEG channels and longer than 1 s have been kept. Noteworthily, it is highly crucial to set the threshold V T so as to get the best performance. Hence, 100 different values of threshold V T , selected from 1 to 100 μV with a step of 1, have been tested.

For each EEG of 11 channels, a vector of 13 features has been extracted as following:

1) the number of IBI, called nb_IBI,

2) the total duration of IBI, which is defined as the sum of all IBI durations, called tot_IBI (seconds),

3) the percentage of IBI in the EEG, called P _ I B I ( % ) = t o t _ I B I E E G _ d u r a t i o n ,

4) the duration of the longest IBI, called Max_IBI (seconds),

5) the maximum of IBI percentage in the EEG, called P _ M a x _ I B I ( % ) = M a x _ I B I E E G _ d u r a t i o n ,

6) the mean duration of IBI which is defined as the sum of the IBI durations divided by the number of IBI, called Mean_IBI (seconds),

7) the number of bursts, called nb_B,

8) the total duration of the bursts that are calculated as the sum of all bursts durations, called tot_B (seconds),

9) the percentage of bursts in the EEG, called P _ B ( % ) = t o t _ B E E G _ d u r a t i o n ,

10) the duration of the longest burst, called Max_B (seconds),

11) the maximum of bursts percentage in the EEG, called P _ M a x _ B ( % ) = M a x _ B E E G _ d u r a t i o n ,

12) the mean duration of the bursts was calculated as the sum of the bursts durations divided by the number of bursts, called Mean_B (seconds),

13) the gestational age of the infant at the time of the EEG examination, called Age_EEG (weeks).

The extracted features and the gestational age form a set of vectors x m ∈ ℝ 13 , m = 1 , ⋯ , M with M the total number of EEG. The entire data set is written as { ( x 1 , y 1 ) , ⋯ , ( x m , y m ) , ⋯ , ( x M , y M ) } with class labels y m ∈ { + 1, − 1 } for Abnormal and Normal EEG respectively. The task hereafter consists of selecting relevant features and discriminating EEG into Abnormal or Normal. Two classifiers were compared: Support Vector Machines (SVM) and Multiple Linear Regressions. In the following, feature selection is explained in the context of both classification methods.

Feature extraction was done along with SVM classification [

Technically, SVM separate the data set { ( x 1 , y 1 ) , ⋯ , ( x m , y m ) , ⋯ , ( x M , y M ) } ∈ ℝ d × { − 1,1 } by a hyperplane with the largest possible margin and the minimal number of misclassified data. This hyperplane is defined by a weight vector w ∈ ℝ d , d being the dimension of feature vectors, and an offset b ∈ ℝ as following:

H : ℝ d → { − 1,1 } x m ↦ s i g n ( w ⋅ x m + b ) (3)

This hyperplane is calculated by solving an optimization problem under constraints:

{ 1 2 ‖ w ‖ 2 + C ∑ m = 1 M ξ m subject to : ( w ⋅ x m + b ) y m ≥ 1 − ξ m ξ m > 0 , ∀ m (4)

1 2 ‖ w ‖ 2 is the maximal margin hyperplane, C is the regularization parameter and ξ m are the nonnegative slack variables [

By setting to zero the derivatives of the partial associated Lagrangian according to the primal variables w , b and ξ m , the optimization problem of the dual formulation can be written as:

{ ∑ m = 1 M α m − 1 2 ∑ m , p = 1 M α m α p y m y p 〈 x m , x p 〉 subject to : 0 ≤ α m and ∑ m = 1 M α m y m = 0 (5)

The linear SVM is extended to a non-linear classifier by mapping data into a higher dimension space using a mapping function Φ , then the optimization problem becomes as follows:

{ ∑ m = 1 M α m − 1 2 ∑ m , p = 1 M α m α p y m y p K ( x m , x p ) subject to : 0 ≤ α m and ∑ m = 1 M α m y m = 0 (6)

where K designs the kernel function. The hyperplane solution has the final following formulation:

h ( x ) = ∑ m = 1 M α m y m K ( x , x m ) + b (7)

Several kernels were tested, namely Radial basis function kernels (RBF), polynomial kernels and linear kernels. As for the dimension d of input data, all combinations of the 13 features were tested for each kernel. This results in testing 2 13 − 1 = 8191 combinations for each kernel and for each of the 100 threshold values aforementioned in 3.

For the implementations, we used Matlab© (The Mathworks Inc., South Natic, MA, USA) and the LS-SVM 1.8 toolbox that provides a complete implementation of SVM [

Multiple linear regression is a generalization of the simple linear regression method [

Y i = a 0 + a 1 X i 1 + a 2 X i 2 + ⋯ + a p X i p + ϵ i , i = 1 , ⋯ , n (8)

where the coefficients a 0 , a 1 , ⋯ , a p are the parameters to be estimated and ϵ i are the errors of the model that expresses the missing informations.

Like for SVM, all combinations of the 13 explanatory variables for each threshold were tested.

To evaluate the accuracy of the predictions, two parameters were used: the sensitivity and the specificity. The percentages of sensitivity and specificity were computed as follows:

• sensitivity = 100 × T P / ( T P + F N ) ,

• specificity = 100 × T N / ( T N + F P ) with:

- TP: number of true positives, TN: number of true negatives,

- FN: number of false negatives, FP: number of false positives.

The use of sensitivities and specificities is based on a precondition: the distribution of “normal” and “abnormal” EEG must be significantly balanced. We reached a prevalence of 11.23%, so this condition of data balance was not met by the corpus of EEG. Therefore, ROC curves were used [

For estimating the generalization error with a small bias and a small variance, we used a K-fold cross-validation [

[ 1 − sensitivity 100 ] 2 + [ 1 − specificity 100 ] 2 (9)

The 5 subsets were built randomly; just keeping an equivalent number of children in each subset: due to the number of 42 abnormal EEG (indivisible by 5), we had 3 sets of 8 abnormal EEG and 2 sets with 9 abnormal EEG.

During the 5 cross validations, 3 kernels (linear, polynomial and gaussian radial basis) were tested. For the polynomial kernel, the degree varied from 3 to 5. The gaussian radial basis worked with σ ∈ [ 0.1 ; 2.0 ] . The optimal SVM kernels (linear, polynomial and gaussian radial basis) that gave the highest mean value of the K AUC were retained.

For linear SVM, the threshold was 35 μV and the selected features were: nb_IBI, P_IBI, P_Max_IBI Mean_IBI, nb_B, P_Max_B. SVM with polynomial kernels reached the optimal performance with a threshold equal to 32 μV using Age_EEG, tot_IBI, P_IBI, Max_IBI, tot_B, P_B, Max_B, P_Max_B, Mean_B. Finally, the gaussian SVM used only 3 features Age_EEG, Mean_IBI, nb_B, with a threshold equal to 25 μV.

The final detector was trained on all the corpus with the Multiple Linear Regression method on the 11 features Age_EEG, nb_IBI, tot_IBI, P_IBI, Max_IBI, P_Max_IBI, Mean_IBI, nb_B, P_B, P_Max_B and Mean_B. With the prediction set to +1 (Abnormal) and −1 (Normal), we obtained the Equation (10) which is detailed in the following.

P = − 0.1936 x 1 − 0.1929 x 2 + 0.1893 x 3 + 0.1246 x 4 + 0.0623 x 5 − 0.0286 x 6 + 0.0104 x 7 − 0.001 x 8 + 0.0007 x 9 − 0.0005 x 10 − 0.0002 x 11 (10)

Method | Sensitivity (%) | Specificity (%) | AUC |
---|---|---|---|

SVM linear kernels | 48.37 ± 13.97 | 98.19 ± 2.23 | 0.73 ± 0.06 |

SVM polynomial kernels | 53.17 ± 11.20 | 97.08 ± 1.63 | 0.75 ± 0.06 |

SVM RBF kernels | 53.17 ± 11.20 | 98.55 ± 1.52 | 0.76 ± 0.05 |

Multiple linear regression | 86.11 ± 10.01 | 77.44 ± 7.62 | 0.82 ± 0.04 |

Variables | Normal (n = 274) | Abnormal (n = 42) | P-value |
---|---|---|---|

nb_IBI | 86.89 ± 67.30 | 175.64 ± 69.01 | P < 0.01 |

tot_IBI (seconds) | 221.01 ± 218.30 | 705.12 ± 414.87 | P < 0.01 |

P_IBI (%) | 11.85 ± 11.70 | 39.90 ± 23.27 | P < 0.01 |

Max_IBI (seconds) | 7.40 ± 5.36 | 22.88 ± 22.35 | P < 0.01 |

P_Max_IBI (%) | 0.40 ± 0.29 | 1.50 ± 2.44 | P < 0.01 |

Mean_IBI (seconds) | 2.06 ± 0.79 | 4.20 ± 3.27 | P < 0.01 |

nb_B | 87.76 ± 67.20 | 176.36 ± 69.00 | P < 0.01 |

tot_B (seconds) | 1627.90 ± 283.28 | 1104.40 ± 465.98 | P < 0.01 |

P_B (%) | 88.15 ± 11.70 | 60.10 ± 23.27 | P < 0.01 |

Max_B (seconds) | 388.40 ± 402.52 | 114.06 ± 177.16 | P < 0.01 |

P_Max_B (%) | 21.63 ± 23.09 | 6.33 ± 9.72 | P < 0.01 |

Mean_B (seconds) | 130.65 ± 357.70 | 9.35 ± 11.61 | P < 0.05 |

where variable P represents the variable prediction, variable x_{1} represents Mean_IBI, variable x_{2} represents nb_IBI,..., variable x_{11} represents Mean_B (all variables are shown in

All calculations were performed on computers equipped with Intel Core i5-3470 CPU at 3.20 GHz, 8 Go of RAM under Linux Ubuntu. We used 10 computers simultaneously: for the 100 thresholds, the linear SVM kernels took 9 days and 14 hours. While the polynomials SVM kernels took 65 days and 8 hours, only 10 days and 2 hours were necessary for the RBF SVM kernels. Finally, the Multiple Linear Regressions took only 59 minutes on one computer.

Experimental results show that a Multiple Linear Regression estimated on 11 features (Age_EEG, nb_IBI, tot_IBI, P_IBI, Max_IBI, P_Max_IBI, Mean_IBI, nb_B, P_B, P_Max_B and Mean_B) can detect accurately abnormal EEG. The detection of an abnormal preterm infant EEG reaches a sensitivity of 95.11% ± 10.01%, a specificity of 77.44% ± 7.62%, and an AUC of 0.82 ± 0.04. Thus, if

Variables | Weight (in %) | Cumulative weight (in %) |
---|---|---|

Mean_IBI | 24.08 | 24.08 |

nb_IBI | 23.99 | 48.06 |

nb_B | 23.54 | 71.60 |

P_Max_IBI | 15.50 | 87.10 |

Age_EEG | 7.75 | 94.85 |

P_B | 3.56 | 98.41 |

P_IBI | 1.29 | 99.70 |

Max_IBI | 0.12 | 99.82 |

tot_IBI | 0.09 | 99.91 |

P_Max_B | 0.06 | 99.97 |

Mean_B | 0.02 | 100.00 |

the automatic detection considers that an EEG is abnormal, it must be interpreted also by the neurologist before undergoing more medical examinations such as an MRI (Magnetic Resonance Imaging). Finally, due to the high sensitivity of our test, an EEG classified as normal does not need to be interpreted urgently by the doctor.

A main advantage of the proposed method is that threshold and feature selection are tuned so as to maximize classification performance. There are of course several ways to select threshold and features [

When comparing SVM to Multiple Linear Regressions, we can see that computational time of linear SVM is 1.32 × 10^{6} times slower, RBF SVM is 1.46 × 10^{6} times slower and polynomial SVM is 9.52 × 10^{6} times slower than that of regressions. Besides, Multiple Linear Regressions performance are higher than SVM ones. However, SVM results are promising, namely those obtained with RBF SVM kernels where only 3 variables were selected (Age_EEG, Mean_IBI, nb_B). This sparsity in feature selection could enhance the robustness of our learning machines [

It is also worthy to note that performances were achieved on a set of 316 EEG after rejecting 100 doubtful EEG. It would be interesting to learn a classifier that could automatically labels these suspicious recordings as ambiguous. The weaknesses of this article relies on the fact that EEG classifications were only achieved by a single EEG expert. This is a major flaw of the proposed system where two or three expert opinions would limit the biases of the predictions. Another limitation of this paper lies in the fact that only SVM and Multiple Linear Regressions were used and not neural networks for example. The reason for this is essentially because it would have taken too long to test all the combinations with neural networks.

This study suggests an automated method to detect abnormal Electroencephalograms (EEG) of preterm infants. The novelty of this paper lies in the combination of these three facts: firstly we work on preterm infants; secondly we propose to automatize the current diagnosis and not to automatize a long term neurological outcome and thirdly this automated prediction is evaluated in a prospective group and not only in a retrospective group. The method consists of detecting Inter Burst Intervals, extracting features from EEG, selecting relevant features and classifying them into normal or abnormal EEG. Thus, gestational age and 10 features (N_IBI, TOT_IBI, P_IBI, MAX_IBI, P_MAX_IBI, MEAN_IBI, N_B, P_B, P_MAX_B, MEAN_B) extracted from the EEG and introduced in a Multiple Linear Regression model, could reliably predict an abnormal finding with a sensitivity of 86.11% ± 10.01%, a specificity of 77.44% ± 7.62% and an AUC of 0.82 ± 0.04.

These results are very promising and encourage further research that could enhance detection of abnormal EEG, namely considering more features, like frequency and information theory features for instance. Finally, testing combination of several classifiers could be a promising path of research too.

This research was paid by no grant. Sincere thanks to J.F. Gelfi and R. Woodward for their help in the improvement of the quality of this paper.

The authors declare no conflicts of interest.

Schang, D., Chauvet, P., Tich, S.N.T., Daya, B., Jrad, N. and Gibaud, M. (2018) Automatic Abnormal Electroencephalograms Detection of Preterm Infants. Journal of Data Analysis and Information Processing, 6, 141-155. https://doi.org/10.4236/jdaip.2018.64009