Seps-is?: A Diagnostic Tool for Infection using Vital Signs

Inspiration

Upon our expeditious (and moderately unpunctual) arrivals to MedHacks 2019, we quickly congregated and decided to form our team of rich and diverse backgrounds. Very soon after, we attended our first track pitch session focused on challenges in post-operative care. We soon learned of the truly intricate difficulties faced in post-operative care, nearly concluding that the broader challenges were insurmountable in the timeframe of a weekend-long hackathon. However, upon a deeper consultation of literature, we collectively concluded that the relative youth of this field was especially interesting. The use of multiple vital signs' data used in tandem as predictive tools for complex diseases was especially inspiring, and we put forth our finest efforts to improve access to predictive models and implementation as at-home diagnostic tools of infection.

What it does

Seps-is? is a demonstrative diagnostic tool for infection using vital sign data inputs. This includes a brief discussion of recent strategies in predicting infection and demonstration of calculations used to predict infection presence (and necessity of medical attention) on randomly generated patient vital sign data. Additionally, the compact character of this program demonstrates the strong potential for implementation in an at-home diagnostic device. The ultimate goal of this project is to demonstrate that clinical predictive models can be employed by non-professional programmers using elementary programming techniques.

How we built it

Python 3.7 was chosen as a simple and comprehensible programming language of choice for this project. The Jupyter Notebook platform was chosen for advantages in compartmentalization of code, data visualization, and distribution.

Challenges we ran into

Lack of appropriate hardware for heart rate and temperature measurements led to challenges in demonstrating the full ability of this predictive model in a microcomputer device.

Accomplishments that we're proud of

A clever project name, of course. Additionally, we were grateful to have put our efforts to targeting challenges that we initially found nearly impossible to address, using direction from current clinical research and simple programming techniques.

What we learned

Fully utilizing our diverse backgrounds in software, hardware, and research to address a clinically-relevant issue with interdisciplinary approaches allowed us to learn a little bit of everything! Additionally, the reduction of biologically complex diseases like infection down to simpler measurements of vital signs demonstrated that complex problems could be addressed with simple approaches.

What's next for Seps–is?

Further research and implementation of predictive approaches for various diseases would significantly advance post-operative care and the medical field overall.

Run the Jupyter Notebook on your own device! Python 3.7 and Jupyter installation required, no additional packages necessary! The static notebook may also be viewed as a PDF or HTML file. See the GitHub link below for the notebook, PDF, and HTML files.