Share
Embed
Healthy Python: Applying Python in Healthcare
Presentation Slides at PyCon PH 2018
https://pycon.python.ph/
Hacarus Inc.
February 24, 2018
Transcript
40 slides
-
Call me Ninz! - Anything Developer @ Hacarus - Backyard
astronomer - Music enthusiast - Physics and astronomy is <3 - I love data! - Curious, always. Github: @pprmint Website: http://ninzz.xyz/ -
Let’s start! • Why healthcare? ◦ Importance of healthcare space
◦ Problems to solve • Why python? ◦ Python strengths ◦ What healthcare problems can be solved by python • How to use python in solving healthcare problems? ◦ Tools ◦ Libraries ◦ Problem solving approach • How to identify effectivity of python in solving these problems? ◦ Effects on stakeholders ◦ Overall development implications • What to expect going forward? This presentation will highlight several key aspects on how we use python in solving healthcare problems and challenges. By the end of the presentation, we should be able to answer the following questions. -
Healthcare is the maintenance or improvement of health via the
prevention, diagnosis, and treatment of disease, illness, injury, and other physical and mental impairments in human beings. (Wikipedia) -
Why Healthcare? - It is one of the (if not
the) most important aspect of our everyday lives - Industry needs innovation and better solutions for existing problems - Healthcare itself is very broad, therefore can be broken down into several smaller industries (more opportunities for ideas and solutions) - Integral part of government. (Although most of them sucks) - Has lots of challenging problems to solve -
Current challenges in healthcare Infrastructural overhaul - Data delivery -
Systems standardization - Difference in procedures and protocols - Data security and data integrity -
Current challenges in healthcare Monetary and operational cost - Affordable
healthcare - Disjoint system and misaligned reward system -
Current challenges in healthcare Personalized healthcare - Consumer applications -
Individual - Organizational - Prevention mechanism - Custom tracking and monitoring - Accessible diagnosis -
Why Python? - Easy to learn and use (but hard
to fully master). - Large library base and support community. - Fast prototyping - Flexibility of use (you can use it for web, scripts, embedded systems, etc) - Scalability -
Solving healthcare problems Python is good in data processing or
anything data related - Data delivery - Systems standardization - Consumer applications - Data security and data integrity -
Solving healthcare problems Strong ML and pattern recognition libraries -
Custom tracking and monitoring - Consumer applications - Accessible diagnosis -
Solving healthcare problems Web technologies in python - Difference in
procedures and protocols - Custom tracking and monitoring - Consumer applications - Affordable healthcare -
Tools Flask - Microframework - Good for fast prototyping -
Flexible structure - Easy to learn and less intimidating - Can scale if structure is done properly -
Tools SQLAlchemy - One of the most reliable ORM tool
for python - Good interface for managing data -
Tools Ansible - Great for automation and task management -
Provides powerful python interface to create programmable automation and configuration setup *As seen on previous presentation* -
Tools Docker - Portable and ideal for deployment - Allows
versioning of images, makes it easy to revert working image -
Tools Jupyter - Great tool for fast visualization and presentation
of data - Easy to use and easy to distribute -
Libraries Boto - Powerful python library to interface with AWS
- Caters different use cases (cmd, imported in files, etc) - Configurable -
Libraries scikit-learn - Robust machine learning python library - Optimized
approach for the different machine learning algorithms - Tested and fast in through its usage in the scientific community in general -
Libraries pandas - Data formatting/data structure library - Ideal for
data analysis - Usually goes hand in hand with other data analysis libraries like scikit-learn -
Libraries Plotly and matplotlib - Visualization library for python -
Works with most data format produced by libraries like pandas and scikit-learn -
Problem solving approach Programming paradigm - Object oriented vs functional
- Depends on what kind of healthcare problem you are solving - I always prefer a mix of both (its not standard, but it’s fast) -
Problem solving approach Architectural Design - Build with a mindset
that data is king. - Must always ensure fast data delivery - Put checks on security and data integrity - Scalable (we use aws load balancers) -
Problem solving approach Machine Learning - Ideal for large data
sets - Can derive patterns in data using several approaches like regression and classification - Lots of supported libraries and tools -
Problem solving approach Sparse Modeling - Learning approach that requires
less data compared to ML and DL - Mostly used in problems related to imaging -
Understanding Stakeholders Aside from being data driven, healthcare startups are
greatly dependent on their stakeholders. It’s not just end users but also other healthcare institutions that you need to consider when jumping into this industry. -
Understanding Stakeholders In order to know if the solutions you
built using python is effective, one must consider the following questions. - Is the data delivery done timely? - Is the data secure? - Is your solution simple and does it target specific niches? - Can your stakeholders identify and use the data delivered by your solution? -
Development Implications You also need to identify if developing your
solutions using python does not entail any technical liability to you and your team. It is not enough to just deliver solutions to the users and healthcare entities. Python can easily be your choice of starting tech, but identifying if using python is still beneficial when moving forward is very important. This part is often overlooked by engineers and leads since they do not like to rewrite/re-engineer their solutions. -
Development Implications Watchout for these: - Sacrificing speed and structure
of your application vs the delivery time. - Pain of adding new features - Difficulty in code refactoring - Difficulty in communicating to outside interfaces -
Healthcare is evolving To be able to succeed in the
healthcare space, it is not enough to build solutions like web app, mobile app, etc. These are disjoint solutions and only adds to the problem of standardization. The best approach would be to build platforms and base infrastructure to which other healthcare solutions can be built on top of. Some of the most successful healthcare startups allow fast communication and interfaces between other providers. -
Healthcare is data driven By now, you should at least
know that healthcare is highly dependent on data. Knowing this, healthcare solutions must evolve to cater the growing need of fast and reliable data delivery. Standardization of data format is also something that needs to be done in order for the healthcare system to work efficiently. This is where python comes in even more. With the advances in data science and data engineering, new approaches and techniques to solve complex data problems should help in building better healthcare systems. -
Some cool resources. . . . - https://medium.com/@mikettownsend/the-top-resources-for-healthcare-entrepreneurs-in- every-category-b35603cac094 -
https://tincture.io/a-guide-to-health-startups-9b4d4f8377b3 - https://medium.com/go-weekly-blog/the-20-most-innovative-companies-in-healthcare-603 8cce9d1f6