Wolfram Summer Camp 2018: Using Machine Learning to Diagnose Pneumonia from Chest X Rays
Author
Rohit Panse
Title
Wolfram Summer Camp 2018: Using Machine Learning to Diagnose Pneumonia from Chest X Rays
Description
Wolfram Summer Camp 2018: Using Machine Learning to Diagnose Pneumonia from Chest X Rays
Category
Essays, Posts & Presentations
Keywords
Wolfram Summer Camp 2018
URL
http://www.notebookarchive.org/2018-12-53xco3y/
DOI
https://notebookarchive.org/2018-12-53xco3y
Date Added
2018-12-11
Date Last Modified
2018-12-11
File Size
1.61 megabytes
Supplements
Rights
Redistribution rights reserved
Download
Open in Wolfram Cloud
WOLFRAM SUMMER CAMP 2018
Last modified on: Thursday, July 12, 2018 at 10:45
Author Info
Name:
Rohit Panse
Mentor:
Michael Kaminsky
End of Camp Presentation Content
Title of project:
Using Machine Learning to Diagnose Pneumonia from Chest X Rays
Goal of the project:
Use a transfer learning with a convolutional neural network to accurately diagnose pneumonia and predict the location if pneumonia is diagnosed.
Add the most representative image of your project here. (We recommend just 1 image, if you add more, we will make a collage of the images.)
Image:
Summary of Results:
I created a transfer learning neural networks which successful diagnosed pneumonia with an accuracy of about 83% and accurately detected the location of the pneumonia with an accuracy of 95%. Data editing and augmentation procedures and increasing the amount of training rounds helped increase accuracy of diagnoses.
Future work:
Improving image formatting function to make the current program more compatible with poor quality chest X-Rays. Using a similar Neural Network set up except with thermal images to efficiently diagnose pneumonia. Further augmenting data to improve accuracy of the neural network.
Detailed Project Notes
Main Results in Detail
Main Results in Detail
Pneumonia is an infection that causes inflammation in one or both of the lungs; it is induced by a variety of organisms: bacteria, viruses, and fungus. Bacterial pneumonia, the more common type, is caused by bacteria that multiplies in the lungs; Viral Pneumonia is caused by an array of viruses, the most common one being Influenza. Although the severity of pneumonia can vary, young children, seniors, and people with a weakened immune system are the most vulnerable.The CDC reports that of the close to 540,000 cases of pneumonia each year, 50,000 people die. Early and proper diagnosis can tremendously decrease mortality rate. A diagnosis is determined by infiltrates, or white spots, present in a chest X-ray. Initially my diagnosing neural network achieved an accuracy rate of about 78%, however after augmenting portions of the data through rotation and perspective shift the accuracy increased to 83%. Four separate neural networks were used to determine what side of the lung had pneumonia; all had an average accuracy of about 95%.
Code
Code
By:Rohit Panse 7/18
By:Rohit Panse 7/18
Part 1 -Diagnosing Pneumonia
Part 1 -Diagnosing Pneumonia
Part 2-Determining Location of Pneumonia
Part 2-Determining Location of Pneumonia
Part 3- Cloud Deploy and Creating a MicroSite
Part 3- Cloud Deploy and Creating a MicroSite
Part 4-Determining Accuracy
Part 4-Determining Accuracy
Cont.
Cont.
Data Sources Links/References
Data Sources Links/References
Background Info Links/References
Background Info Links/References
Cite this as: Rohit Panse, "Wolfram Summer Camp 2018: Using Machine Learning to Diagnose Pneumonia from Chest X Rays" from the Notebook Archive (2018), https://notebookarchive.org/2018-12-53xco3y
Download
