At the forefront of Artificial Intelligence

Home » Articles » Neural Networks » Applications

Neural Networks in Anaemia Classification

1.0 Introduction

The pioneering work of neural network in the modern era has started since 1943 by McCulloch and Pitts. To date, there has been an explosive growth research in this field and has attracted many investigators, including academician, physicians, psychologist, and neurobiologist beginning early 1980s. An approach to the pattern recognition problem was introduced by Rosenblatt (1958) in his work on the perceptron and they are now many successful projects and on-going projects that utilized the ability of neural networks in their applications.

The applications of neural networks are almost limitless but they fall into several main categories like classification, modeling, forecasting and novelty detection. Some examples of successful applications include; credit card fault detection (Alaskerov et al., 1997), pattern recognition (Ramli et al., 1996, Rietveld et al., 1999), handwritten character recognition (Le Chun et al, 1990; Tay & Khalid, 1997; Karim et al., 1998), colour recognition (Yaakob et al., 1999), and share price prediction system (Sanugi et al., 1996; Lim et al., 1996) and others. Many researchers have compared Artificial Neural Netwoks (ANNs) and Logistic Regression (LR) models. They have shown that neural networks are able to make a better generalization over the traditional statistical methods such as regression techniques (Lapuerta et al., 1995; Shanker, 1996; Lapuerta et al.; 1997, Armoni, 1998).

2.0 Neural Networks in Medical

One of the major goals of observational studies in medicine is to identify patterns in complex data sets. Literatures have shown that medical has benefited much from this technology. It has been successfully applied to various areas of medicine to solve non-linear problems. The applications include prediction of diagnosis such as cancers (Astion et al., 1992; Wilding et al, 1994), the onset of diabetes melitus (Shanker, 1996), survival prediction in AIDs (Ohno-Machado, 1996), eating disorders (Buscema et al., 1998) and others. Applications in signal processing and interpretation involve EEGs or electroencephalogram analysis (Makeigh et al., 1996), ECGs or electrocardiograms (Bortolan et al., 1991), EMGs or electromyelogram (Chiou et al., 1994), and EGGs or electrogastrograms classifications (Lin et al., 1997).

Performance of the neural network strategy has shown higher performance than Cox regression models in predicting clinical outcomes of the risk of coronary artery disease (Lapuerta et al., 1995). In addition to this study, Lapuerta et al. compared the prediction of survival of neural networks and logistic regression models on alcoholic patients with severe liver disease. The study reveals that neural networks were more successful in classifying patients into low and high-risk group.

A similar study carried out by Armoni (1998) shows that neural network prediction was more accurate than linear regressions for prediction the diagnostic probabilities of insulin-dependent diabetes mellitus. The results suggest the use of a neural network should be considered whenever prediction of diagnosis is required. In the area of medical image processing, Doffner et al. (1996), demonstrated that neural network can be effectively be used as a tool in medical decision-making. They applied neural network in the interpretation of planar thallium-201 scintigrams for the assessment of coronary artery disease.

3.0 Application in Haematology

There is an attempt to create an expert system to diagnose classes of anaemia and report presumptive diagnoses directly on the haematology form (Birndort et al., 1996). The purpose is to simulate the processes of human experts that can reliably achieve diagnostic separability by pattern analysis. In doing this, they constructed a hybrid expert system combining rule-based and artificial neural network (ANN) models to evaluate microcytic anaemia in a 3-layered program using haematocrit (HCT), mean corpuscular volume (MCV), and coefficient of variation of cell distribution width (RDWcv) as inputs. These measurements are available as standard output on most haematology analyzers. Three categories of microcytic anaemia were considered, iron deficiency (IDA), haemoglobinopathy (HEM), and anaemia of chronic disease (ACD). The performance of the model was evaluated with actual case data. The results show that the model was successful in correctly classifying 96.5% of 473 documented cases of microcytic anaemia and anaemia of chronic disease. This result exhibits sufficient accuracy to be considered for use in reporting microcytic anaemia diagnoses on haematology forms.

The leukocyte-vessel wall interactions are studied in post capillary vessels by intravital video microscopy during in vivo animal experiments (Egmont-Petersen et al., 2000). Sequences of video images are obtained and digitized with a frame grabber. A method for automatic detection and characterization of leukocytes in the video images is developed. Individual leukocytes are detected using a neural network that is trained with synthetic leukocyte images generated using a novel stochastic model. This model makes it feasible to generate images of leukocytes with different shapes and sizes under various lighting conditions. Experiments indicate that neural networks trained with the synthetic leukocyte images perform better than networks trained with images of manually detected leukocytes. The best performing neural network trained with synthetic leukocyte images resulted in an 18% larger area under the ROC curve than the best performing neural network trained with manually detected leukocytes.

4.0 Anaemia Classification

"Anaemia" is a common medical problem. The word anaemia is composed of two Greek roots that together mean "without blood" (Ed-Uthman, 1998). Signs and symptoms of anaemia include weakness fatigue, palpitation, light-headedness, difficulty in swallowing, loss of appetite, nausea, constipation, diarrhea, stomatitis and others. The patient looks pale, the nail may be dry and brittle, and tongue may be inflamed. In severe anaemia, heart failure and swelling of both limbs can occur. In mild anaemia, none of the above signs and symptoms may appear (Orkin, 1992). According to him, patients with anaemia have a significant reduction in red cell mass and a corresponding decrease in the oxygen-carrying capacity of the blood. In General Medical Officer's manual (Luiken et al., 1999), anaemia is define as a decreased level of haemoglobin more than two standard deviations below the expected mean for age and sex. Anaemia itself is not a disease but a sign of disease (Rapaport, 1987; DeLoughery, 1999). This means underlying disease is presents that demand an explanation.

This paper presents an empirical evaluation on medical, particularly classification of anaemia patients using neural networks approach. The number of hidden units, learning rate and momentum are varied, so that more appropriate classification model is obtained. The information regarding anaemia cases were collected from haematology form which was used in the government hospital. Seventeen attributes were indentified and used in the training model. A total of seven hundred raw data of anaemia patients has been collected and preprocessed that includes data cleansing, data selection and data preprocessing. Data is partitioned into three data sets namely training set (80%), testing set (10%) and validation (10%) set. The training data is used to train the model while the validation data is used to monitor neural network performance during training. The test data is used to measure the performance of a trained model. The motivation here is to validate the model on a data set that is different from the one used for parameter estimation. Figure 1 shows the schematic diagram of multilayer perceptron with with 17 units of input layers, 15 units of hidden units and 8 units of outputs.

Figure 1: Schematic Representation of the model

5.0 Discussion and Conclusion

The highest performance was obtained when the number of hidden units is 15, learning rate is 0.7 and momentum is 0.1. The testing and generalization correctness is 71.56 and 72.78 respectively. This result has demonstrated the ability of multilayer perceptron for predicting classes of anemia and can be used by haematologist and other medical staff.

For future work, an improvement can be tackled in the following aspects to intensify the usefulness and the generalization of the model. In the pre-processing phase for example, the proper method was not used to remove the unequal distribution of the data set. The data was randomly removed using the ordinary statistical software. The data that has been removed may contain important information. A better method such as hierarchical neural networks (HNNs) as suggested by Ohno-Machado (Ohno-Mahado, 1996a) can provide a way of enhancing the sensitivity to rare categories without decreasing its specificity.

References

Alaskerov, E., Freisleben, B. and Rao, B., 1997. CARDWATCH: A Neural Network Based Data Mining System for Credit Card Fraud Detection, Proceedings of the IEEE/IAFE, 220-226.

Armoni, A., 1998. Use of neural networks in medical diagnosis, MD Computing, Mac-Apr;15(2):100-4

Astion, M.L., and Wilding, P. 1992. Application of neural networks to the interpretation of laboratory data in cancer diagnosis, Clin Cehm Jan;38(1):34-8

Birndorf, N.I., Pentecost, J.O., Coakley, J.R., and Spackman, K.A. 1996. An expert system to diagnose anemia and report results directly on hematology forms, Comput Biomed Res Feb;29(1):16-26.

Bortolan, G., et al. 1991. "ECG classification with neural networks and cluster analysis." Proceedings Computers in Cardiology. Held: Venice, Italy, 23-26 Sept.

Buscema, M., Mazzeti, di P.M., Salvemini, V., Inraligi, M., and Indrimi, M. 1998. Application of artificial neural networks to eating disorders. Subst Use Misuse Feb:33(3):765-91.

Chiou, Y.S., and Lure, Y.M. 1994. Hybrid lung nodule detection (HLND) system. Cancer Letters, Mar 15, 77(2-3):119-26.

DeLoughery G.T., 1999, Anemia: An Approach to Diagnosis, Division of Hematology & Medical Oncology, March 15.
http://www.ohsu.edu/som-hemonc/handouts/deloughery/anemia.shtml

Dorffner, G., Erich, P., Markus, M., Stefan, K., Paolo, P., Gerold, P., Heinz, S., 1996. Experiences with Neural Networks as a Diagnostic Tool in Medical Image Processing.

Doi, K., Giger, M.L., Nishikawa, R.M., and Schmidt, R.A. 1997. Computer vision and artificial intelligence in mammography: Proceedings of the International Symposium on"Diagnosis and Therapy of Breast Cancer", February, Germany (in press).

Ed Uthman, MD, 1998, Understanding Anemia,University Press of Mississippi;
http://www.neosoft.com/~uthman/unanemia/unanemia_outline.html

Egmont-Petersen, M., Schreiner, U., Tromp, S.C., Lehmann, T.M., Slaaf, D.W. 2000. Detection of leukocytes in contact with the vessel wall from in vivo microscope recordings using a neural network, Arts T Department of Biophysics, Maastricht University, The Netherlands. IEEE Trans Biomed Eng, Jul;47(7):941-951

Karim, M.E., Hossain, M., Mottalib, M.A. , and Zaman, T. , 1998. A modified Neural Network Learning Approach and Its Application to Bengali Character Recognition, MJCS Volume 11, No. 2.9.

Lapuerta, P., Azen, S.P., and La, B. 1995. Use of Neural Networks in predicting the risk of coronary artery disease, Comput Biomed Res, Feb;28(!):38-52

Lapuerta, P., Rajan, S., and Bonacini, M. 1997. Neural networks as predictors of outcomes in alcoholic patients with severe liver disease, Hepatology. Feb;25(2):302-6

LeCun, Y., 1993. Efficient Learning and Second-order Methods, A Tutorial at NIPS 93, Denver.

LeCun, Y., Boser, B., Denker, J.S., Henderson, D., Howard, R.E. Hubbard, W., and Jackerl, L.D. 1990, Handwritten digit recognition with a backpropagation network. Advances in Neural Information Processing Systems, 2, 248-257.

Lim, S.H., Liang, W.T., and Subramanian, R.K. 1996. Neural Network Based Share Price Prediction System, ATIP96.067 : Malaysian Computer Science & Applications (REDECS '96), Universiti Putra Malaysia, 26-27 June.

Lin, Z., Maris, J., Hermans, L., Vandewalle, J., and Chen, J.D. 1997. Classification of normal and abnormal electrogastrograms using multilayer feedfordward neural networks, Med Biol Eng Comput, May; 35(3):199-206

Luiken, G., Phares, J.C., and Millard, F. 1999. General Medical Officer (GMO) Manual: Clinical Section: Hematology. http://www.vnh.org/GMO/ClinicalSection/26Aenmia.html

Makeigh, S., Jung, Tzyy-Ping., and Sejnowski, T.J. 1996. Using Feedforward Neural Networks to Monitor Alertness from Changes in EEG Correlation and Coherence, Advances in Neural Information Processing Systems 8, MIT Press, Cambridge MA, 931-937.

Ohno-Machado, L., 1996. Sequential use of neural networks for survival prediction in AIDs, Proc AMIA Fall Symp, 170-4.

Orkin, 1992. Harrison's Principle of Internal Medicine, Mc-Graw Hill, 13th Edition: Vol1, 313

Ramli, A.R., and Kumar S.V., 1996. Pattern Recognition for IC Mark Inspection, ATIP96.067 : Malaysian Computer Science & Applications (REDECS '96), Universiti Putra Malaysia, 26-27 June.

Rapaport, S.I., 1987. Introduction to Hematology, 2nd Edition, J.B. Lippincott Company, Philadelphia,10.

Reggia, J.A., 1993. Neural Computation in medicine. Artificial Intelligence in Medicine, Apr, 5(2):143-57.
Rietveld, S., Oud,M., and Dooijes, E.H. 1999. Classification of asthmatic breath sounds: preliminary results of the classifying capacity of human examiners versus artificial neural networks. Computers and biomedical research, Faculty of Psychology, University of Amsterdam, The Netherlands, Comput-Biomed Res, Oct, Vol. 32(5) 440-8.
Rosenblatt, F., 1958. The Perceptron: A probabilistic model for information storage and organization in the brain, Physiological Review, Vol. 65, 386-408.

Sanugi, B., and San, W.S., 1996. Neural Network Approach In Predicting KLSE Composite Index, LT/M.Bil. 7.

Shanker, M.S., 1996. Using Neural Networks to Predict the Onset of Diabetes Mellitus, Department of Administrative Sciences, Kent State University, Kent, OH 44242, In Journal of Chemical Information and Computer Sciences, 36

Tay, Y.H., and Khalid, M. 1997. Comparison of Fuzzy ARTMAP And MLP Neural Networks For Hand-written Character Recognition, IFAC Symposium on AI in Real-Time Control 1997 (AIRTC'97), Kuala Lumpur, Malaysia, 363-371.

Yaakob, R., Sulaiman, M.N., Mahmood, R., Mohamed, M.T.M., and Ramli, A.R, 1999. Multi-Counterpropagation Network Model For Colour Recognition, MJCS, Vol. 12:1.5.

Wilding, P., Morgan, M.A., Grygotis, A.e. Shoffner, M.A., and Rosato, E.F. 1994. Diagnosis of backpropagation neural networks to diagnosis of breast and ovarian cancer, Cancer Lett Mar 15:77(2-3):145-53

Submitted: 19/03/2004

Article Toolbar

BibTeX entry