aravindmathradan / sih19_ezdi

Inconsistency in Medical Annotations

Project Source Code can be downloaded from the github repository ykaran86/SIH19_ezDI

Authors

1. Yogesh Karan

email:- [email protected]

2. Disha Dudhal

email:- [email protected]

3. Devasish Mahato

email:- [email protected]

4. Dhanashree Revagade:-

email:- [email protected]

5. Shivam Dev Singh

email:- [email protected]

6. Aravind Unnikrishnan

email:- [email protected]

Supervisor

Yesoda Bhargava

https://yesodabhargava.com/

Contributors

Dhanashree Revagade

Shivam Dev Singh

Devasish Mahato

Disha Dudhal

Aravind Unnikrishnan

created and tested on Ubuntu 18.04 with Python 3.6

An Attempt(which finally led us to the victory) to provide a solution to the company ezDI in the complicated type problem statement Identify Inconsistency in Medical Annotations in the Smart India Hackathon(Software Edition) 2019. This is a Web App developed in Django which displays the clusters of semantically similar medical terms having inconsistencies in Annotations. This project is based on strict dictionary lookup based approach. Technologies used in this project:

1. Django (Web Framework)
2. QuickUMLS python module for looking into the UMLS
3. UMLS for providing medical insights
    (https://www.nlm.nih.gov/research/umls/licensedcontent/umlsknowledgesources.html)

Input to the application:- .tsv file containing sentences and their annotations

Output of the application:- .tsv file containing medical groups with their annotation patterns (apart from the web App)

Features provided in the Web App:

1. Clusters with unique annotation patterns are displayed 
  (in order such that clusters with maximum inconsistency are shown first)
2. On Clicking a particular pattern of a cluster,  sentences containing that pattern are shown.
3. Search operations are also enabled so that one can search for occurence of 
    a particular word in the provided dataset or a particular annotation tag
4. Statistical Insights are also provided which displays the proportion of inconsistency 
    in the data or contribution of multi-word entities vs single word entities in the inconsistencies

Front Page of the Web App

On Uploading a .tsv file

Uploading will take time as in this step preprocessing and clustering is done!

Home Page of the Web App

Search Page

On searching a word

Here, sentences containing blood are displayed

On searching a tag

Here, sentences with problematic temrs are displayed

On searching a part of a word

Even on searching a part of a word, sentences are displayed

The Clusters with Unique Annotation Patterns

On clicking a pattern sentences containing that annotation pattern are displayed

Inconsistency vs Consistency

Inconsistency in Single Word Entities vs Multi Word Entities

Few Words to Visitors

This Application is in its development phase and its obvious to get erroneous cluster patterns in some cases. A lot of improvement is needed. Pull requests for any such changes are accepted. Feel free to fork this project and make your own changes too.

Thank you for Visiting!