DIAGNOMATIC: DISEASE DETECTION AND VISUALIZATION USING MEDICAL IMAGING AND DEEP LEARNING

Abstract

Pneumonia is particularly a difficult disease to be diagnosed because it could be caused by any number of pathogens that lead to a bacterial, fungal or viral infection in the lungs and it can be contracted almost anywhere, including in hospitals. Chest X-rays have been considered the best tool to detect any form of pneumonia, but it's not perfect. Pneumonia can appear similar to other conditions on scan, and imaging cannot identify the infectious pathogens, making the diagnosis difficult via x-ray. Physicians must gather a complete picture of the patient's condition to make their best and most accurate diagnosis. The work portrayed targets automating entire procedure of diagnosing the Pneumonia and visualizing it using a web-app with a high accuracy of 92% which could make it easier for medical practitioners and patients to understand the area diseased in the given X-ray. The model uses X-ray images to predict the actual diseased location by the use of widely used Deep Learning Algorithms such as CNN (Convolutional Neural Network) and Data Augmentation. The deep learning model is trained using 5,863 JPEG images extracted from retrospective cohorts of pediatric patients of one to five years old from X-Ray imaging was performed as part of patients' routine clinical care. The model is trained using the Kaggle’s cloud kernel and then saved in H5 and json format which is later used by the Django framework to generate output is the newly inputted images into the web-app by end user via frontend that is implemented using the vue.js framework. This project uses python 3.x libraries TensorFlow and Keras to implement Data Augmentation and CNN Algorithms. To perform other visualization, mathematical and, statistical operations it used NumPy, Pandas, Seaborn, Scikit Learn and OpenCV libraries. Other than these it uses Django and Vue.js to implement Backend and frontend of the web-app respectively.

Dependency

1. Install Virtual Environment pip install virtualenv

2. Create Virtual Environment python -m venv environment

3. Activate the Environment /environment/Scripts/activate

4. Install Django pip install django

5. Install Django REST Framework (For Creting APIs) pip install django-rest-framework

6. Install Django Cors for creating TCP/IP Headers for API calls pip install django-cors-headers

7. Install Djoser for Authentication pip install djoser

8. Install Pillow (PIL) for Image Processing pip install pillow

9. Install Strip pip install stripe

10. Create a Django Project django-admin startproject diagnomatic_django

11. Configuring the Django Project a. Configuring settings.py INSTALLED_APPS = [ ... ... 'rest_framework', 'rest_framework.authtoken', 'corsheaders', 'djoser', ]

    CORS_ALLOWED_ORIGINS = [
    	"http://localhost:8000", # Where the site is allowed to run
    ]
    
    MIDDLEWARE = [
    	'django.middleware.security.SecurityMiddleware',
    	'django.contrib.sessions.middleware.SessionMiddleware',
    	'corsheaders.middleware.CorsMiddleware',
    	'django.middleware.common.CommonMiddleware',
    	'django.middleware.csrf.CsrfViewMiddleware',
    	'django.contrib.auth.middleware.AuthenticationMiddleware',
    	'django.contrib.messages.middleware.MessageMiddleware',
    	'django.middleware.clickjacking.XFrameOptionsMiddleware',
    ]
    

    b. Configuring urls.py urlpatterns = [ ... path('api/v1', include('djoser.urls')), path('api/v1', include('djoser.urls.authtoken')), ]

12. Migrate Database python manage.py migrate

13. Create Superuser python manage.py createsuperuser username: diagnomatic email: [email protected] password: *123dani#

14. Setup Frontend (Vue.js) a. Install Node.js b. Install Vue Cli npm install -g @vue/cli c. Create Vue Project vue create diagnomatic_vue => Configuration Manually i) Babel, Router, Vuex, CSS-Preprocessor ii) 3.x (Preview) iii) History mode for router -> Yes iv) CSS-Preprocessor -> Sass/SCSS (with dart-sass) v) deidicated config Files