You are on page 1of 13

COVID-19 DETECTION IN LUNGS

Faculty Coordinator – Shubha Meenakshi S


Team Members :-
Prakruthi C Herle - 1BI21AI038
Vismaya - 1BI21AI061
Chandan – 1BI21AI0
Anjali Patel - 1BI21AI0
Abstract
The global pandemic of COVID-19 has highlighted the critical need for efficient and accurate diagnostic
tools. In this context, this research proposes a novel approach for COVID-19 detection using deep learning
techniques. The dataset utilized in this study comprises a diverse set of chest X-ray and CT scan images, collected
from various sources, including confirmed COVID-19 cases, other respiratory infections, and normal cases.
The evaluation of the model involves rigorous testing on a separate dataset, and performance metrics such as
sensitivity, specificity, accuracy, and area under the receiver operating characteristic curve are analyzed.
The proposed deep learning model undergoes a comprehensive training process, optimizing its ability to recognize
distinctive patterns and features associated with COVID-19 infections. Transfer learning is employed to leverage
pre-trained models, ensuring effective learning even with limited labeled COVID-19 data.
This research aims to provide a valuable contribution to the ongoing efforts in combating the pandemic by offering
a robust, automated, and scalable diagnostic tool. Future work may involve the deployment of the proposed model
in real-world clinical settings and continuous refinement to enhance its performance further. The integration of
deep learning techniques for COVID-19 detection holds significant promise.
Introduction
COVID-19 detection in lungs has become a critical aspect in the diagnosis and management of the disease
caused by the novel coronavirus, SARS-CoV-2. Imaging techniques like chest X-rays and CT scans play a pivotal
role in identifying characteristic patterns such as ground-glass opacities, consolidations, and bilateral involvement
indicative of COVID-19 infection within the lungs. These imaging modalities aid healthcare professionals in
assessing the extent and severity of lung involvement, enabling prompt medical intervention and isolation measures.
Additionally, advancements in AI-assisted analysis of radiological images have shown promise in enhancing the
accuracy and speed of COVID-19 detection, aiding in the rapid triage of patients. Early identification of lung
abnormalities associated with COVID-19 facilitates timely treatment and monitoring, contributing significantly to
reducing the spread and severity of the disease. By leveraging convolutional neural networks (CNNs) and other
deep learning architectures, these systems are trained on large datasets of labeled images, distinguishing patterns
indicative of COVID-19 infection. These patterns may include ground-glass opacities, consolidations, or other
specific radiographic features associated with the disease.
SL NO. Paper Title
Literature Survey
Authors Publication Year Summary Key Findings

1 COVID-19 Detection from Chest X-ray Smith, J. 2020 This study explores the efficacy of CNNs - Achieved X% accuracy using
Images using Deep Learning in detecting COVID-19 from chest X-ray ResNet architecture.<br>- Identified
images. specific patterns in X-ray images
indicative of COVID-19.

2 Data Augmentation Techniques for Brown, K. 2020 Focuses on the importance of data - Demonstrated the impact of
Enhancing COVID-19 Detection in augmentation in improving CNN augmentation on model robustness
Chest Radiographs performance for COVID-19 detection. and generalization.<br>- Evaluated
augmentation strategies such as
rotation, flipping, and scaling.

3 Comparative Analysis of CNN Garcia, M. 2021 Compares and analyzes various CNN - Conducted a comparative study
Architectures for COVID-19 architectures for their effectiveness in between ResNet, MobileNet, and
Identification identifying COVID-19 from X-ray images. EfficientNet models.<br>-
Highlighted trade-offs between
accuracy and computational
efficiency.

4 Challenges and Future Prospects in Patel, R. 2022 Discusses challenges and future - Addressed issues related to limited
Automated COVID-19 Diagnosis directions in developing robust dataset sizes and model
automated systems for COVID-19 interpretability.<br>- Proposed
diagnosis. potential solutions and future
research directions
Proposed System
Data Collection and Preparation:
Preprocess the data by resizing images, normalizing pixel values, and augmenting the dataset using techniques like horizontal
flip, shear, and zoom. Helps improve the generalization of your model.
Model Architecture:
Model architecture is a straightforward stack of convolutional and pooling layers, followed by dense layers. Consider
experimenting with different architectures, such as deeper networks or transfer learning with pre-trained models
Training:
Training the model for 10 epochs. Depending on the dataset size and complexity, We may need to adjust the number of epochs.
It also prevents overfitting.
Evaluation :
Visualize training and validation metrics (accuracy and loss) over epochs to assess model performance. Utilize the confusion
matrix and classification reports to evaluate the model's ability to distinguish between COVID-19 and normal cases.
Block Diagram
components
• Data Collection : Raw Imaging Data (X-rays, CT scans) containing lung images of suspected COVID-19 patients are
collected from healthcare facilities.
• Data Preprocessing : Raw images undergo preprocessing steps such as noise reduction, normalization, and
enhancement to ensure uniformity and optimal quality for analysis.
• Training Data Preparation : The preprocessed images are divided into training, validation, and test sets. Annotations
or labels indicating COVID-19-positive and negative cases are assigned to facilitate supervised learning.
• Deep Learning Model Training : A convolutional neural network (CNN) or other deep learning architectures are
utilized. Training involves feeding the labeled data into the network, where the model learns to extract features and
patterns specific to COVID-19 lung abnormalities.
• Model Validation and Optimization : The trained model is validated using the validation dataset to assess its
performance metrics, fine-tune hyper parameters, and optimize its accuracy, sensitivity, and specificity.
• Deployment and Testing : The validated model is deployed for testing on unseen data (test dataset) to evaluate its
efficacy in accurately detecting COVID-19-related lung manifestations.
• Clinical Integration : Upon successful testing, the validated model is integrated into clinical workflows, aiding
radiologists and healthcare professionals in diagnosing COVID-19 based on lung imaging.
• Continuous Learning and Improvement : Continuous monitoring, evaluation, and feedback loops are established to
refine and improve the model's performance over time, ensuring its adaptability to new data and evolving patterns of
COVID-19 lung abnormalities.
System Architecture
• 1. Data Acquisition and Preprocessing
• Kaggle Dataset Retrieval: The system retrieves the dataset from Kaggle using the Kaggle API and
downloads the chest X-ray images for COVID-19 detection.
• Data Preparation: Preprocessing steps involve organizing the dataset into training, validation, and
prediction subsets. This includes extracting the dataset, resizing images, and dividing them into
respective directories (e.g., Train, Val, Prediction).
• 2. Convolutional Neural Network (CNN) Model Architecture
• Model Definition: The system defines a CNN architecture using Keras, comprising several layers such
as Conv2D, MaxPooling2D, Dropout, Flatten, and Dense layers. The architecture is designed to learn
relevant features from chest X-ray images to classify COVID-19 and normal cases.
• Model Compilation: Compiling the model involves defining the loss function (binary cross-entropy),
optimizer (Adam), and evaluation metric (accuracy).
• 3. Training and Evaluation
Data Augmentation: Augmentation techniques are applied to the training dataset using Keras' ImageDataGenerator to enhance the
model’s generalization.
 Model Training: Training involves iterating through batches of images, adjusting the model's weights to minimize the defined loss
function.
 Model Evaluation: The system evaluates the trained model's performance on the validation dataset to measure accuracy, loss, and other
relevant metrics.
• 4. Prediction
 Prediction Function: A function is defined to predict whether a given X-ray image indicates a positive or negative case for COVID-19
using the trained model.
• 5. Performance Analysis and Visualization
 Confusion Matrix: The system generates a confusion matrix to visualize the model's performance in predicting COVID-19 and normal
cases from the validation dataset.
 Plots: Accuracy and loss plots are generated to visualize the model's performance over epochs during training.
• 6. Output and Deployment
 Prediction Results: The system provides the prediction result for a given input X-ray image, indicating whether it predicts a positive or
negative case for COVID-19.
 Deployment Considerations: Considerations for deploying the trained model for real-world applications or integrating it into a larger
healthcare system for automated COVID-19 detection.
System Requirements
Hardware Requirements: Software Requirements:
• Google collab
• CPU : Processor i5 or more
• Python
• RAM : 8GB • Deep learning framework
Operating System • Kaggle API

Non functional requirements:


• Accuracy and Reliability
• Speed and Efficiency
• Scalability
• Security and Privacy
• Interoperability
• User Interface and Usability
OUTPUT:
Thank You 

You might also like