Training Neural Networks as Experimental Models: Classifying Biomedical Datasets for Sickle Cell Disease

Abstract

Abstract. This paper presents the use of various type of neural network architectures for the classification of medical data. Extensive research has indicated that neural networks generate significant improvements when used for the pre-processing of medical time-series data signals and have assisted in obtaining high accuracy in the classification of medical data. Up to date, most of hospitals and healthcare sectors in the United Kingdom are using manual approach for analysing patient input for sickle cell disease, which depends on clinician's experience that can lead to time consuming and stress to patents. The results obtained from a range of models during our experiments have shown that the proposed Back-propagation trained feed-forward neural network classifier generated significantly better outcomes over the other range of classifiers. Using the ROC curve, experiments results showed the following outcomes for our models, in order of best to worst: Back-propagation trained feed-forward neural net classifier: 0.989, Functional Link neural Network: 0.972, in comparison to the Radial basis neural Network Classifiers with areas of 0.875, and the Voted Perception classifier: 0.766. A Linear Neural Network was used as baseline classifier to illustrate the importance of the previous models, producing an area of 0.849, followed by a random guessing model with an area of 0.524.

Improving pain management in patients with sickle cell disease from physiological measures using machine learning techniques

“…In paper (Desai et al, 2012), a machine learning algorithm based on support vector machines was adopted to identify a 10-gene signature that discriminates between patients with and without increased tricuspid regurgitation jet velocity (TRV), and validated it as a potential biomarker for an elevated TRV in SCD. Khalaf et al (2016) presented various neural network models for classifying the level of dosage for SCD medication. They used 13 features, including body weight, Hemoglobin, and Mean Corpuscular Volume to recommend one out of 6 levels of hydroxyurea medication dosage the patient needs to take, and obtained the best performance with AUC of 0.989.…”

Applying Neural Network-Based Approach to Sickle Cell Disease-Related Pain Classification

“…This will help improve the patient's quality of life with attendant reduction of unnecessary spending, patient illness and pressure for the healthcare practitioners in terms of emergency cases they need to attend to per time. The SCD attributes adapted from Khalaf et al (2016) in Table 4 were enhanced for consideration with Age, Educational Background and Location items. Edeki and Akanbi (2017) used Monte Carlos Simulation (MCS) Technique to complement the Physical Simulation Smith's Statistical (PSSS) package in simulating data with respect to Sickle Cell Anaemia (SCA) in order to examine the mathematical inheritance formation of the SCA disease.…”

Machine Learning Techniques for Automatic Detection of Sickle Cell Anemia using Adaptive Thresholding and Contour-based Segmentation Method

“…[22][23][24] Other popular methods proved very useful are, namely, watershed transform, [10] improved watershed algorithm, [8,25] marker controlled watershed technique, [26] and circular Hough transform (CHT). [27][28][29][30][31][32] In recent years, application of NN in classification and detection of medical images has increased significantly because of its high accuracy and less computation time. [29,30] Khalaf et al presented a work on the utilization of different types of models of neural networks for detecting biomedical dataset for sickle cell disease.…”

“…They considered four different types of NN approaches: Feed-forward NN, functional link NN, radial basis NN and voted perception classifier, and applied for classification to determine medication dosage for SCD patient. [31] Elsalamony proposed an algorithm for detecting anemia causing abnormal RBCs from microscopic images with the help of CHT and morphological operations. After that, using NN the resulting data of detection process is analyzed.…”