How and why I built Web Application for Closed-monitoring patients
About me
My passion for technology has always been a driving force in learning new things and apply the same in expanding my knowledge in the field. Later, I developed my passion more deeply through undergraduate studies in software engineering from Delhi Technological University, one of the prestigious universities in India. I am a hard-working student, and I will commit myself to effectively take advantage of the opportunity given to me. The feeling that my work has a direct impact on people's lives motivates me.
The problem I wanted to solve
- Taking care of our loved ones could be overwhelming. Sometimes it becomes difficult to find a balance between working at office and taking care of our parents and children. What's even worse is to pay for loved ones, 63% of Indians dip into their own savings and that's not including the money you lose from taking time off work.
- In case of nursing homes, the amount of risk associated is higher. Most of them don’t even have doctors available on site or on call. Infact, only 72 hours of training is given to the nurses which is inadequate.
- This case is no different when comes to pre-schools, there's no accurate way to track and find what your little one is upto during his/her hours at school for a parent who wants to be a part of their day, all they get are vague responses.
- In case of physically challenged users, closed monitoring like tracking location and vitals to alert concerned users when required becomes essential.
What is Web Application for Closed-monitoring patients?
- This problem is indeterministic in nature in our opinion we can only approach it, not completely solve it, and we are doing through multiple heuristics.
- With CareWheel, we are building a platform for users (like children, senior citizens, physically challenged people) and caretakers (nursing homes, primary schools, teachers, etc).
- A small-affordable IoT device (depending on your type of supervision) will keep monitoring a user's vitals including - (health related) temperature, heart rate, etc (over 37 parameters) as well as (general parameters) location, outside temperature, humidity.
- Caretakers can upload status of individuals like images and video logs.
- All this information is sent to our backend where we use over 20 different statistical techniques to sort the data to get more meaningful insights along with Machine Learning techniques to predict recovery rate & mood of the user.
- We're reducing the paperwork done by caretakers and automating the process making it more transparent and reliable for user's guardians (ones who are monitoring the user).
- Our platform can remind caretakers about user's medicine schedule or food patterns, etc which otherwise is a task .
- Caretakers have their personal web-dashboard where they can set medicine schedule, specify boundaries which when crossed by user will alarm the concerned person (guardian).
- Finally, using NLP our whatsapp chat bot is powerful enough to provide all the relevant information to the concerned guardian of the user in a familiar interface without having to learn some new-technology.
Tech stack
IOT:
- Internet of Things (IoT) measures and reports data by blending in our daily lives. There are wide variety of data we can collect from them and it will vary from user to user.
- Currently, we have used Arduino to collect ultrasonic distance (range 2-200 cm) for closed monitoring, heart beat sensor, and GPS sensor.
Whatsapp Chat Bot:
- It allows the guardian to get updates about the individual (elderly or child) directly from the backend.
WhatsApp Chatbot is used because of its ease of use and high familiarity to the users. - Using NLP (Natural Language Processing) queries need not to be very specific because our system is capable of understanding the semantics.
Flask Backend & Machine Learning:
- Our Flask backend is highly scalable and maintainable.
- IoT devices along with caretakers (through web-dashboard) directly sends data to the backend using REST API which can be further assessed through whatsapp chat bot or web interface.
- Our Smart decision service is hosted in the cloud which can: Calculate recovery rate, Alert concerned people whenever a new trend is detected (like sudden rise of heart rate), Periodically update statistical charts to provide updated information, Calculate mood from text & image posts
Artificial Intelligence:
- Using Random Forest model we train the following dataset and predict whether the person is recovering or not based on the data provided. It gives a confidence of 87% and also decreases overfitting and is successful in handling different variety of datasets (missing values & outliers). Other techniques fail in this regard by a margin of 10-20% as shown below.
- Using multimodal deep learning we are doing semantic analysis with following dataset: https://nlp.stanford.edu/projects/glove/
- Overall the trained model is giving a confidence of 62.37%
The process of building Web Application for Closed-monitoring patients
Challenges I faced
Key learnings
- Our approach has a highly intuitive user interface and unburdens the caretaker and the guardian from learning a new mode of data collection.
- Flask REST-API backend, provides room for scalability (IoT devices), collaboration, and data accessibility.
- Real-time notification keeps the track of unnatural trends and gives instant feedbacks in case of emergencies.
- Entire patient report is accessible through the web app.
- Currently, fitness bands are used to sync data to your phones but with CareWheel we are taking it a step further by creating a platform where both users and supervisors come together such that no crucial information goes waste.
