Cardiotocogram improve classification using the neural network classification system

Susan Casoori

Abstract


Cardiotocography (CTG), consisting of fetal heart rate (FHR) and tocographic (TOCO) measurements, is
used to evaluate fetal well-being. It is one of the most common diagnostic techniques to evaluate maternal and fetal
well-being  during  pregnancy  and  before  delivery.  By  observing  the  Cardiotocography  trace  patterns  doctors  can
understand the state of the fetus. Even few decades after the introduction of cardiotocography into clinical practice,
the predictive capacity of the existing methods remains inaccurate.  In a previous work (Sundar.C and et al, 2012),
we showed that a model based CTG data classification system using a supervised artificial neural network (ANN)
can classify the CTG data better than most of the other methods. But, the performance of the normal neural network
based classifier was limited because of the presence of potential outliers in the training data. The presence of outliers
in  training  data  affects  the  neural  network  training  as  well  as  testing.  In  this  work,  we  present  improved
classification  models  which  will  consider  outliers  in  the  data  and  eliminate  them  from  training  phase  of  the
classification process. We used Precision, Recall, F-Score and Rand Index as the metric to evaluate the performance.
The proposed idea considerably improved the performance in  classifying Normal, Suspicious and Pathologic CTG
patterns.  It was found that, the improved classifier was capable of identifying Normal, Suspicious and Pathologic
condition with very good accuracy than normal methods.

References


The following chart shows the performance of BL-BPN algorithm. In general, the algorithm gives good

performance for normal and pathological records and poor performance in suspicious records.

Classification Performance of BL-BPN

93

96 0.95

71

67

69

90

70

76

00

10

20

30

40

50

60

70

80

90

00

Precision Recall F-Measure

Metric

Score .

Normal Suspicious Pathological

Figure 10. Performance of BL-BPN algorithm

The derived results obviously show that the proposed bi-level training improved the classification

performance of system. The BL-BPN approach provided good performance in all cases than compared other

methods.

Conclusion

We have evaluated the performance of the four methods with respect to three different metrics. The

performance of standard neural network based classification model, RBF, and SVM were has been compared with

proposed BL-BPN Model. According to the derived results, the performance of the proposed supervised machine

learning based classification approach provided significant performance than other compared methods.

It was found that, the proposed BL-BPN based classifier was capable of identifying Normal, Suspicious

and Pathologic condition, from the nature of CTG data with very good accuracy. If we see the performance of BL-

BPN with respect to all the metrics, then we can say that it almost provided double the performance of the other

three compared methods.

So, future works may address the way to improve the system to recognize the suspicious CTG patterns and treat

them separately while training and testing. One may address the way to improve the system for getting proper

training with different classes of CTG patterns. Future works may address hybrid models using statistical and

machine learning techniques for improved classification accuracy.

References

Xiaojun Chen, Yunming Ye, Xiaofei Xu, Joshua Zhexue Huang , “A feature group weighting method for

subspace clustering of high-dimensional data”, Pattern Recognition 45 (2012) 434-446, Elsevier

Shahad Nidhal, M. A. Mohd. Ali1 and Hind Najah, “A novel Cardiotocography fetal heart rate baseline

estimation algorithm”, Scientific Research and Essays Vol. 5(24), pp. 4002-4010, 18 December, 2010

Ana. Klimesova, Eva Ocelikova Multidimensional Data Classification, Proceedings of the 10th WSEAS

International Conference on Automation & Information, ISSN: 1790-5117, ISBN: 978-960-474-064-2

Stirrat, Mills and Draycott, "Notes on Obstetrics and Gynaecology for the MRCOG, 5th Edition", 04 Aug

, ISBN: 9780443072239

Diogo Ayres-de-Camposa, Cristina Costa-Santosb, Joa Bernardesa, "Prediction of neonatal state by computer

analysis of fetal heart rate tracings: the antepartum arm of the SisPorto1 multicentre validation study”,

European Journal of Obstetrics & Gynecology and Reproductive Biology 118 (2005) 52-60.

Shomona Gracia Jacob and R. Geetha Ramani, "Evolving Efficient Classification Rules from

Cardiotocography Data through Data Mining Methods and Techniques", European Journal of Scientific

Research, ISSN 1450-216X Vol.78 No.3 (2012), pp.468-480

Sundar.C, M.Chitradevi and Dr.G.Geetharamani “Classification of Cardiotocogram Data using Neural

Network based Machine Learning Technique” International Journal of Computer Applications, ISSN 0975 – 888 Vol. 47 No. 14 (2012) pp. 19-25.

Han, Jun; Morag, Claudio (1995). "The influence of the sigmoid function parameters on the speed of

backpropagation learning". In Mira, José; Sandoval, Francisco. From Natural to Artificial Neural Computation.

p. 195-201

Sundar.C, M.Chitradevi and G.Geetharamani, “An Analysis on the Performance of k-means Clustering

algorithm for Cardiotocogram Data clustering” International Journal on Computational Sciences &

Applications, ISSN 2200 – 0011, Vol. 2 No. 5 (2012) pp. 11-20.

http://www.academicjournals.org/SRE, ISSN 1992 – 2248 , 2010 Academic Journals

Antonia Costa, MD; Diogo Ayres-de-Campos, PhD; Fernada Costa, MD; Cristina Santos, MS; Joao Bernardes,

PhD, “Prediction of neonatal academia by Computer analysis of fetal heart rate and ST event sibnals” 2009

AJOG – American Journal of Obstetrics and Gynecology.

http://www-h.eng.cam.ac.uk/help/tpl/ programs/ matlab.html.

Barnett, V. and Lewis, T.: 1994, Outliers in Statistical Data. John Wiley & Sons., 3rd edition.

http://www.iasri.res.in/ebook/win_school_aa/notes/Evaluation_Measures.pdf

Sundar Chinnasamy, Chitradevi Muthusamy and Geetha Ramani G “ An Analysis on the Performance of

Fuzzy C -Means Clustering Algorithm for Cardiotocogram Data Clustering” Caspian Journal of Applied

Sciences Research, ISSN 2251 – 9114 Vol. 1 no. 13 pp. 35-42, 2012

C. Domeniconi, D. Gunopulos, S. Ma, B. Yan, M. Al-Razgan, D. Papadopoulos, 2007.Locally adaptive metrics

for clustering high dimensional data, Data Mining and Knowledge Discovery 14 (1) 63–97.

K. Cios, W. Pedrycz, and R. Swiniarsk, 1998. Data Mining Methods for Knowledge Discovery. Boston:

Kluwer Academic Publishers.

J. Han and M. Kamber, 2000. Data Mining; Concepts and Techniques, Morgan Kaufmann Publishers

Ben Kao, Sau Dan Lee, Foris K.F.Lee, David W. Cheung, Wai-Shing Ho,” Clustering Uncertain Data using

Voronoi Diagrams and R-Tree Index” IEEE Transactions on Knowledge and Data Engineering, Vol. 22(9),

pp. 1219 – 1233, sep 2010.

Bin Jiang; Jian Pei, 2011, Outlier detection on uncertain data: Objects, instances, and inferences, IEEE 27th

International Conference on Data Engineering, ,doi: 10.1109/ICDE.2011.5767850

Onisko, A., & Druzdzel, M. J. 2011. Impact of quality of Bayesian network parameters on accuracy of medical

diagnostic systems. In Working Notes of the 2011 AIME’11 Workshop on Probabilistic Problem Solving in

Biomedicine. Bled, Slovenia.

Philip A Warrick, Emily F Hamilton, 2011. Subspace detection of the impulse response function from intra-

partum Cardiotocography, Conf Proc IEEE Eng Med Biol Soc. 2011 Aug, pp5678-81

Chen CY, Chen JC, Yu C, Lin CW.2009. A comparative study of a new Cardiotocography analysis program.

Conf Proc IEEE Eng Med Biol Soc. 2567-70.


Refbacks

  • There are currently no refbacks.