Inspiration
Sepsis a life-threatening condition that occurs when the body’s response to an infection causes tissue damage, organ failure, and eventually death. Globally, nearly 30 million people develop sepsis and 6 million people die from sepsis each year. Over one third of people who die in US hospitals die from sepsis. Sepsis costs US hospitals more than any other condition at $24 billion (13% of US healthcare expenses). Every hour of delayed treatment has been associated with roughly a 4-8% increase in morality. As it is difficult to tell if a patient will experience septic shock, as the condition moves quickly and looks like a normal infection, predictive analytics through artificial neural networks, a process that rapidly processes relationships complexly which doctors cannot observe, will allow physicians to achieve a greater understanding of a patient's chances of developing sepsis
What it does
Our model takes in patient data from up to the last 20 timestamps and rapidly assesses the empirics to produce a prediction of whether the patient will experience sepsis. This gives doctors a relatively accurate and quick prediction that analyzes many variables to aid their diagnosis, allowing them to know which patients need to be monitored closely.
How we built it
We parse pre-existing clinical data from two different hospitals, preprocessing data with class weights, Stratified Shuffle Split, and more so that the classes of sepsis and no sepsis are equally represented and assessed in our algorithm. We feed this newly processed data into the artificial neural network, which contains three layers with 5, 3, and 2 nodes respectively, so the computer can analyze the relationships between the variables and produce our predictive learning model. We also utilized pycharm and html to produce the website, which is designed to take patient data and feed that into our machine learning model.
Challenges we ran into
First of all, the data was massively skewed, only 6% of the data was of septic patients. In order In order to sync the machine learning model with the website, we had to utilize pycharm (which we weren't familiar with before) and quickly learn the language before doing so.
Accomplishments that we're proud of
We're proud of learning pycharm on the spot, as well as implementing a functioning machine learning algorithm in a short period of time.
What we learned
We learned pycharm as well as various techniques to make sure the classes in our classification model aren't skewed, like class weights and stratified shuffle splits. We learned how to integrate machine learning algorithms into a website.
What's next for Diagno-sepsis
We will better Diagno-sepsis by making our product more user friendly by allowing users to input values individually. We will also provide a login function to protect patient data, as well as integrate the system to preexisting medical database to make it more accessible to users, such as doctors.
Log in or sign up for Devpost to join the conversation.