Sie sind auf Seite 1von 22

EASWARI ENGINEERING COLLEGE

Department of Information Technology

First Review

Malaria Diagnosis with Peripheral Blood Smear Images using


Deep Convolutional Neural Networks

Project by,
Supervisor Shri Dikshanya K N(310615205082)
Dr.M.Mohana M.E.,Ph.D., B.Shruthi(310615205083)
Associate professor
V.Vinothini(310615205099)
❖ Title OUTLINE
❖Abstract
❖Introduction
❖Objective
❖Existing System
❖Literature survey
❖Limitations in existing system
❖Proposed system
❖Architecture
❖List of Modules
❖Algorithm
❖Experimental Results
❖Performance Analysis
❖Pseudocode
❖References
ABSTRACT
❖ Malaria is a life-threatening disease caused by bites of infected female
anopheles mosquito.
❖ The most widely recognized approach to analyze malaria is by physically
looking at blood smears for parasite-affected red blood cells
❖ This technique is tedious and diagnosis relies on the experience of the
individual performing the examination.
❖ Image processing methods based on machine learning have been applied to
malaria blood smears for diagnosis previously.
❖ However, the practical performance has not been effective so far.
❖ The proposed system provides a novel method to automatically classify the
cells present in human peripheral blood smear images as either infected or
uninfected based on Convolutional Neural Network (CNN).
❖ The simple architecture of the proposed achieved equally good results in
comparison to more complex architectures.
❖ The model could be deployed for simple medical applications that support
manual cell identification
INTRODUCTION

• Identification of Plasmodium parasites in a blood sample using a


microscope is seen as the best way for laboratory confirmation of malaria.
• However, the malaria screening done manually the whole day is a tedious
process. The analysis is very time-consuming and might even lead to
imprecision and deviation in some cases.
• The automated advancement in malaria detection uses microscopic images
of blood slides captured from high resolution camera that will be able to
improve consistency in malaria diagnosis.
• Computer Aided Diagnosis (CAD) of malarial parasite is performed by first
segmenting Red Blood Corpuscles.
• The proposed algorithm uses a deep learning approach for classification of
the blood smear images into infected and uninfected.
• Thus it helps in overhauling the current laboratory technology into
computerized.
OBJECTIVE

• To use Deep Convolutional Neural Networks to automatically classify the


cells present in human peripheral blood smear images as either infected or
uninfected
EXISTING SYSTEM

❖ BLOOD SMEAR PROCESS


The blood smear process can be visualized here:

Figure 1 : Blood Smear Process


• First, a blood sample is taken from a patient and then placed on a slide.
• The sample is stained with a contrasting agent to help highlight malaria
parasites in red blood cells.
• A clinician then examines the slide under a microscope and manually
counts the number of red blood cells that are infected.
• According to the official WHO malaria parasite counting protocol, a
clinician may have to manually count up to 5,000 cells, an extremely
tedious and time-consuming process.
EXISTING SYSTEM

❖ RAPID DIAGNOSIS TESTING


An example of an RDT device used for malaria testing can be seen below:

Figure 2 : Rapid Diagnosis Testing


• Here you can see a small device that allows both a blood sample and a
buffer to be added.
• Internally, the device performs the test and provides the results.
• While RDTs are significantly faster than cell counting but they are also
much less accurate.
LITERATURE SURVEY
 Malaria Parasite Detection from Peripheral Blood Smear
Images Using Deep Belief Networks.
• This paper uses DBN based model to classify images into parasite or non-
parasite class.
• Restricted Boltzmann machine is used to train DBN using contrastive
divergence method.
• Color and texture is used as feature vector in this paper

 Machine learning approach for automated screening of malaria


parasite using light microscopic images
• In this paper, Microscopic images are used for image acquisition from
stained slides.
• Illumination correction and noise reduction, erythrocyte segmentation,
feature extraction, feature selection was done.

• Finally classification of different stages of malaria have been investigated.


• Bayesian learning and support vector machine (SVM) was used to facilitate
classification.
LITERATURE SURVEY
❖ Segmentation of erythrocytes infected with malaria parasites
for the diagnosis using microscopy imaging.

• An edge-based segmentation of RBCs infected with malaria parasites using


microscopic images has been developed to facilitate the diagnostic process.

• Fuzzy C-means clustering is applied to extract infected erythrocytes, which


is further processed for the final segmentation.

❖ Automated image processing method for the diagnosis and


classification of malaria on thin blood smears

• Image features based on color, texture and the geometry of the cells as well
as features that make use of a priori knowledge of the classification
problem and mimic features used by human technicians.

• A two-stage tree classifier using backpropagation feedforward neural


networks distinguishes between true and false positives, and then diagnoses
the species of the disease.
LIMITATIONS OF EXISTING
SYSTEM

 Requires segmentation of parasite-affected cells


 Optimization of parameters is not possible for big dataset .
 Initialization makes the training process computationally expensive,
resulting in requiring graphical processing units for model training.

 More time is required to train and test the dataset.


 It considers only the image intensity thereby producing unsatisfactory
results in noisy images.

 With edge detection techniques, high quality cannot be achieved for


images which have inherent speckle noise and textural characteristics.
PROPOSED SYSTEM

• In the proposed system, first the pathologists-curated image dataset for


training deep neural network is compiled
• This is followed by pre-processing techniques such as data augmentation
methods used to significantly increase the size of the dataset to avoid
overfitting problem associated with training deep convolutional neural
networks.
• CNNs eliminate the need for manual feature extraction, so you do not
need to identify features used to classify images.
• The features obtained from the training process are stored in separate
array. These feature vectors can be taken as input for classification
purpose.
• We will then compare the classification accuracies obtained by deep
convolutional neural networks through training, validating, and testing
with various combinations of the datasets.
ARCHITECTURE
LIST OF MODULES

 IMAGE ACQUISTION
 PRE-PROCESSING
 FEATURE EXTRACTION
 CLASSIFICATION
IMAGE ACQUISITION

• These images were contributed by Jane Hung of MIT and the Broad
Institute in Cambridge, MA.

Figure 3 : Malaria Dataset

• Images are in .png or .jpg format. There are three sets of images
consisting of 1364 images (~80,000 cells) with different researchers
having prepared each one: from Brazil (Stefanie Lopes), from Southeast
Asia (Benoit Malleret), and time course (Gabriel Rangel).

• Blood smears were stained with Giemsa reagent.


PRE PROCESSING

 Pre-Processing
• In image processing, the initial is the pre-processing stage.
• Improving image quality and image size changes are done in the pre-
processing stage.
• The image that is being the object mostly has poor quality, Through this
initial pre-processing operation, the image quality is improved so that the
image can be used for further stages.

Figure 4: Image Augmentation


FEATURE
EXTRACTION
• CNNs eliminate the need for manual feature extraction, so you do not
need to identify features used to classify images.
• The CNN works by extracting features directly from images.

Figure 5 : Features extracted from the image


• The relevant features are not pre-trained.
• They are learned while the network trains on a collection of images.
• This automated feature extraction makes deep learning models highly
accurate for computer vision tasks such as object classification.
CLASSIFICATION
 Classification
• The features obtained from the training process are stored in separate
array. These feature vectors can be taken as input for classification
purpose.
• The Convolutional neural network is one of the deep learning mechanisms
mainly used for classification, recognition and detection purpose.

Figure 6: Le-Net-5 Architecture


• The input features have been fed, and the corresponding target classes are
assigned.
• The comparison will start between the database features and the testing
features , finally the matched target class is retrieved in the output layer
from that we can able to classify the uninfected or infected blood smear
images.
ALGORITHM

Input : NORMAL/ABNORMAL BLOOD SMEAR IMAGES

1: Sliding Window Process


2: fe ← Extract features from the image
3: repeat:
4: Forward Propagation:
5: cd ← Convolution2D(fe);
6: mp ←max_pooling(cd);
7: fc ←Fully_connected(mp);
8: class label ← soft_max(fc);
9: Backward Propagation:
10: conduct backward propagation with adam;
11: Until wi Convergences;
12: Use the trained network to predict the labels.
ALGORITHM
EXPERIMENTAL
RESULTS
PERFORMANCE
ANALYSIS
THANK YOU

Das könnte Ihnen auch gefallen