M

MedHub_360

MAKING HEALTH CARE EASY AND ACCESSIBLE FROM ANYWHERE

Created on 10th December 2020

M

MedHub_360

MAKING HEALTH CARE EASY AND ACCESSIBLE FROM ANYWHERE

The problem MedHub_360 solves

Challenges in Indian Medical System :
Most of the medical records we have today are in unstructured form. For a patient, it is difficult to store tons of data in the form of doctor’s notes, discharge statements, prescriptions, and lab reports which are all on the paper.
Such data is mostly unstructured and very difficult to analyze.
The past medical data is important for a doctor because in our country, we go to various doctors depending on their specialties and the current doctor must know what all has been administered in the past for better diagnosis.
On top of that, because of the obvious difficulties of storing them as hard copies, some of them might be misplaced which may make it all the more difficult for a doctor to help.
By using this application, the doctor doesn't have to sift through tons of paperwork and the data is available in a structured way which makes the diagnosis fast and effective. With the Symptoms Checker System, many people in our country who don't have access to nearby emergency medical facilities can check with the help of this application by what disease they may be suffering from.

Our Goals :
Extract, store, and retrieve medical information from Medical Reports using OCR.
Display the information online Graphically for the use of Patients as well as Doctors only with Authorised Access.
Disease prediction from existing medical records using ML-Models, to Caution Patients and Recommend them for Check-Ups.
Giving a complete health Analytics to doctors of Patients as well as Doctors preserving user privacy.

Challenges we ran into

Challenges:-
Extracting important information from medical reports using Tesseract OCR and integration with backend servers.
Maintaining user privacy so that private data is not leaked.
Understanding and Rendering Apex Charts.
Training and Deploying a Deep Learning Model.

Solution:-
After trying with several medical reports we found a particular format to extract the data from medical reports.
To maintaining user privacy we used hashing to hash the data, used JSON Web Tokens to prevent unauthorized access.
and also giving uniqueID to every user after registrations.
The size of the Tensorflow trained model was quite huge hence deploying it into Heroku created issues.
We used different file formats(.pkl) and older versions of TensorFlow to deploy.

Discussion

Builders also viewed

See more projects on Devfolio