Web Application for Intracranial Hemorrhage Detection

The objective of this project is to create a web tool for neuro-specialists to assist them in brain hemorrhage detection and localization task. An abstract of this project has been submitted to the Society for Imaging Informatics in Medicine 2019.

Update: This project abstract has been accepted at the SIIM Conference! Congratulations to everyone who worked, guided and supported us.

Table of Contents

• Directory Layout
• How to Run
  • Requirements
  • Steps
• Sample DICOMS
• Demo
• Courtesy
• Contributors
• Supervisors

Directory layout

├── app.py
├── demo
├── Sample DICOM Images
├── gradcam
├── inference_rsna.py
├── model
├── models
│   └── fold0_ep3.pt
├── README.md
├── requirements.txt
├── static
├── templates
├── test.py
├── uploads
└── utils.py

Requirements

• Python packages: Available in requirements.txt
• ngrok

Steps

• Set Up ngrok:

  • Open an account using your mail address here
  • Find your authtoken here.
  • Make ngrok executable on your system and run the following command:

  $ ./ngrok authtoken <YOUR_AUTH_TOKEN>
  

• Running Web server:

  • Go to Intracranial_Hemorrhage_Web_App directory.
  • Run python app.py. The web server will start running on localhost:9999
  • Open another terminal and run ngrok http 9999. After a couple of seconds, it will generate a texts that will look like this

ngrok by @inconshreveable                                       (Ctrl+C to quit)
                                                                                
Session Status                online                                            
Account                       <your_name> (Plan: Free)                               
Update                        update available (version 2.3.35, Ctrl-U to update
Version                       2.3.34                                            
Region                        United States (us)                                
Web Interface                 http://127.0.0.1:4040                             
Forwarding                    http://********.ngrok.io -> http://localhost:9999 
Forwarding                    https://********.ngrok.io -> http://localhost:9999
                                                                                
Connections                   ttl     opn     rt1     rt5     p50     p90       
                              243     0       0.00    0.00    0.01    3.81    

• Copy the link after Forwarding (http://********.ngrok.io). Now anyone can access the web server from this link.

Sample DICOMS

Sample DICOM Images directory contains 12 images. Their labels are available in the Ground_Truth.csv files.

Demo

• Step step tutorial for using the web application
• Demo Video

Courtesy

• 11th place solution of the RSNA Hemorrhage Detection Competition.
• Grad-CAM ++ Pytorch

Contributors

• Tahsin Mostafiz
• Shajib Ghosh

Supervisors

• Dr. Taufiq Hasan
• Dr. Paul Naggy