Detection Of Malaria Disease Using Convolutional Neural Network

downloadDownload
  • Words 2024
  • Pages 4
Download PDF

Abstract:

In the modern era, people are affected by different types of health issues; one of them is Malaria which may be incipient among all aged people. Malaria is a major disease which may be infected by female mosquito bite and it is spread from one people to another by using mosquitoes. The traditional mechanism to detecting malaria disease is visually to examine the blood smears for identifying red blood cells which are affected by malaria-parasites under the microscope in the appearance of an experienced technician. This method is inefficient due to the absence of lab equipment and diagnosis is dependent on the seniority or experience of the person. The main intention of this study is to identify the presence or absence of malaria parasites. To overcome the problems in the traditional malaria detecting system, we develop an automated system to detect malaria parasites. For this implementation, we referred to the Kaggle dataset, which consists of 27,558 images that belong to two classes. In this research work, we use a deep learning algorithm for identifying the presence of the malaria parasites in the blood smears of the people.

Keywords:

Blood smears, CNN, Deep learning, Malaria disease, Microscope, Parasites.

Click to get a unique essay

Our writers can write you a new plagiarism-free essay on any topic

I. Introduction

Malaria is a transmittable disease caused by the parasites which belong to the Plasmodium family. Malaria disease can be spread by the bite of the female mosquito. Every year around 228 million people are affected by malaria around the globe. The number of deaths due to malaria disease is almost 4,05,000. Most of the people who are affected by malaria are children aged below 5 years. The number of children who are died because of malaria disease is 67% (2,72,000) around the globe. The Africa region consists the high malaria cases and high death rates.

In general, malaria symptoms are two types: they are Uncomplicated and severe.

Uncomplicated malaria Symptoms are last 6 to 10 hours and they may reoccur every second day. Uncomplicated malaria consists the symptoms like cold, hot and sweating together with the development of the symptoms. The symptoms development is given below:

  • Feeling of cold accompanied by shakings.
  • Fever, headaches, and vomiting.
  • Annexation sometimes occurs in younger people with the disease.
  • Sweats come after by a return to normal temperature, with exhaustion.

In severe malaria, laboratory confirmation shows the symptoms of organ dysfunction. Symptoms of severe malaria are:

  • Fever and shivering.
  • Unconsciousness.
  • Weakness, or acquire a prone position.
  • Number of attacks.
  • Deep breathing and respiratory distress.
  • Unusual blood loss and symptoms of anemia.
  • Clinical icterus and proof of essential organ dysfunction.

Malaria disease can be easily identified by using the image processing technique along with the deep learning algorithm. It is necessary to develop an automated system for detecting the disease. This study aims to make an automated system that can find the parasites of malaria in blood smears using CNN. The system performance is evaluated based upon accuracy.

II. Related work

Numbers of researches are done to detect the malaria parasites on several datasets using some Machine learning algorithms.

Saiprasath G, Naren Babu R applied machine learning algorithms to find the malaria parasites [1]. For this work, they used a 2703 images dataset which is collected from the Makerere laboratory images database. They use Machine learning algorithms to detect the disease. Among all the algorithms Random Forest gives better performance than the other algorithms. The accuracy of the Random Forest algorithm is 96.5%.

Research done by Sneha Narayan Chavan [2] used two machine learning algorithms to find out the parasites of malaria disease and analyze the malaria disease using Image processing. They use the SVM classifier and NN classifier for their research. In the proposed method, they used 30 images that are personally taken from laboratories. This study intends to differentiate the infected from uninfected malaria cases using slide images of blood smear. The SVM classifier gives 98.25% accuracy while the NN classifier gives 78.53% accuracy for this dataset.

Ahmedelmubarak Bashir, Zeinab A.Mustafa had done their research to detect malaria parasites [3]. For this work, they used Artificial Neural Network (ANN) along with the image processing methods. For this research, they have used 77 images for database creation. In this, the performance of the classification method was evaluated using sensitivity, specificity, and accuracy. For this data set, they got the sensitivity 100%, specificity 99.65% and achieve 99.6% accuracy.

The research was done by Edy Victor Haryanto S, M. Y. Mashor [4] to detect malaria by using Giemsa stain. In this research, they follow some of the methods of image processing; they use this method for conversion of RGB image to grayscale and image screening to remove noise and to improve the quality of the image. The next process is the image segmentation. Then the next step is feature extraction. For this, they use the K-means algorithm. By using this they build an automated system for detecting malaria disease.

III. Proposed work

Image Processing: Image processing is a technique to execute some actions on images in the process of getting features of images or to get the quality image. Image processing is similar to signal processing. Methods of image processing are two types, one is analog and the other one is digital. The image processing which is used for hard copies is known as Analog image processing and another is used for digital images by using a computer is known as Digital image processing.

Deep Learning: Deep learning (DL) is an artificial intelligence method that works as a brain function of humans. DL is a part of machine learning which consists of the algorithms capable of learning both supervised and unsupervised data. Deep learning is also named as a deep neural learning or deep neural network. Deep learning consists of more number of layers used to get higher-level features from the raw input. Most of the deep learning algorithms consisting of the structure or architecture of neural networks, because of this deep learning model are often referred to as deep neural networks.

The most popular classification algorithm of Deep Learning for classification of images is “Convolutional Neural Networks (CNN or ConvNet)”.

CNN is of the deep learning technique that is used for classifying the images. CNN takes an image as input and set learnable weights and take out features or objects in the image and used to differentiate one from another one. It referred to as ConvNet. The working rule of the ConvNet is to reduce the images into a form that is easier for processing.

CNN build-up of and number of hidden layers, but only one input and one output layer. The CNN hidden layers consist series of several layers which are listed below.

  1. Convolution layer.
  2. Pooling layer.
  3. Relu layer.
  4. Fully connected layer.

Convolution Layer: This is the first layer in the architecture of CNN which is used to get features from the image. Convolution of an image with various filters can perform operations like edge detection, sharpening.

Pooling Layer: The use of this layer is to train the algorithm to take a few numbers of parameters of an image when it is too large. They are three types. One is Maximum pooling which takes the highest element from the rectified feature map. Another one is Average pooling which takes the average of all the elements from the rectified feature map and another is Sum pooling which takes all elements sum.

ReLU Layer: ReLU stands for the Rectified Linear Unit. It is an activation function used for the outputs of CNN neurons. Mathematically, it is defined as y = max (0, x).

Fully Connected Layer: Fully connected (FC) layer is a feature vector for input. The operation of the FC layer is to flatten the high-level features which are learnable by Convolutional layers and combining features. It passes the flattened output which is generated to the output layer.

The architecture shown below Fig 9 describes the steps which are used when we develop a system that is used to detect the parasites of malaria from the blood smears.

Implementation process of malaria detection system.

  1. The first and foremost step of this procedure is the data collection. For this work, we collected data that is already available on the internet. The dataset is collected from the Kaggle which consists of 27,558 images. These images are divided into two classes named as parasitized and uninfected. Both of these classes consist of 13779 images.
  2. After data collection, the next step is data preprocessing. In the preprocessing step, we convert all the images into 30*30 dimensions. Then the next step is data labeling. In this step, we represent the parasitized images as 1 and uninfected images as 0 for the data.
  3. The next step is data splitting. In this step, data is divided into two different parts which are known as train data and test data. In this process, the splitting ratio of the data set is (80, 20) i.e. 80% of the data is used for training data and the remaining 20% of data is used for testing.
  4. After data is splitting into train and test then we choose the CNN algorithm for classifying the data. We give the train data to the CNN algorithm by making use of the fit function and train the algorithm to detect the malaria disease. For the training process, we use the 22046 images of the taken data.
  5. After training the next process is testing the data. In this process, we have to test the given data to whether the algorithm classifies the given data correctly or not and shows the predicted class label for the dataset. For this process, we use the 5512 images of the taken data for testing purposes.

IV. Performance Metrics Analysis

Performance metrics are used to differentiate between the algorithms. In this, we use the “accuracy, precision, recall, f1-score, and support” as performance metrics. Formulas for Performance metrics are given below.

From the above formulas, TP stands for true positive, TN stands for true negative, FP stands for False-positive and FN stands for False Negative.

Confusion Matrix

A confusion matrix is a function to measuring the algorithm performance. It is also known as the “error matrix”. It is a table that shows the performance of the classification algorithm. In this, the row denotes the actual class label and the column denotes the predicted class label. The confusion matrix structure is shown below.

V. Results

Malaria Cell Image dataset consists of 27559 images. We divided the dataset into an 80-20 split ratio i.e. 80% (22046) images for training and 20% (5512) images for testing. Precision, recall, and f1-score achieved using Convolutional Neural Network are shown below.

VI. Conclusion

We have proposed a Malaria disease identification system using a deep learning technique. This method of detecting malaria disease can be very efficient for detecting the disease within less time and with fewer errors. This automated system is useful in where there are fewer experts and a lack of resources. In this proposed work, the Convolutional Neural Network algorithm is applied to the Malaria cell image data set for identifying the malaria parasites in blood samples of humans and achieved an accuracy of about 96%.

Our system for the automatic malaria diagnosis is provided useful for disease detection. Furthermore, our automated system framework can be used for the detection of other diseases like tuberculosis, cancer, etc.

References

  1. Saiprasath G, Naren Babu R, ArunPriyan J, Vinayakumar R, Sowmya V, Soman K P, “Performance Comparison Of Machine Learning Algorithms For Malaria Detection Using Microscopic Images” in www.ijrar.org (E-ISSN 2348-1269, P- ISSN 2349-5138).
  2. Sneha Narayan Chavan, Ashok Manikchand Sutkar, “Malaria Disease Identification And Analysis Using Image Processing”, International Journal of Latest Trends in Engineering and Technology (IJLTET).
  3. Ahmedelmubarak Bashir, Zeinab A.Mustafa, Islah Abdelhameid, Rimaz Ibrahem, “Detection of Malaria Parasites Using Digital Image Processing”, International Conference on Communication, Control 2017, Computing and Electronics Engineering (ICCCCEE) Khartoum, Sudan.
  4. Edy Victor Haryanto S, M. Y. Mashor, A.S. Abdul Nasir and Zeehaida Mohamed “Identification of Giemsa Staind of Malaria Using K-Means Clustering Segmentation Technique” The 6th International Conference on Cyber and IT Service Management (CITSM 2018) Inna Parapat Hotel – Medan, August 7-9, 2018.
  5. Naveen Erumalla, et al., “Detection of Malaria Parasite By Blood Smear Examination And Antigen Detection: A Comparative Study”. International Journal of Medical Research & Health Sciences. Volume 2 Issue 1 Jan-Mar 2013.
  6. Gonzalez, Rafael C, and Woods, Richard E and Eddins, Steven L,.2009, “Digital image processing using MATLAB. Gatesmark Publishing Knoxville”.
  7. Mori F, Mori T. Region segmentation and object extraction based on virtual edge and global features[C]//Computer Vision-ACCV 2012 Workshops. Springer Berlin Heidelberg, 2013, pp.182-193.

image

We use cookies to give you the best experience possible. By continuing we’ll assume you board with our cookie policy.