Our Work - AI Model Improves Ulcer Treatment Decision-Making
AI Facilitates Ulcer Treatment Decision Making
Health professionals apply their expertise to determine the appropriate procedure to treat ulcers using their knowledge of both the patients and wounds. The decision-making process is complex given the quantity of available treatment options and the quantity of patient and wound characteristics that must be considered.
Best Option turned to Wovenware to run an innovation sprint and create a proof of concept for and AI predictive algorithm that could help health professionals with this decision process.
What We Faced
- Work with a limited dataset of 73 unique patients and 402 wounds and having only positive data, where treatment was effective but none where treatment was not effective or was not needed.
- Build a predictive model for key treatment options including Primary Bandages, Secondary Bandages and Negative Pressure.
- Build a recommender system for appropriate wound treatment in a limited timeframe.
How We Helped
To address, Best Option’s goal was to validate the feasibility of improving the efficiency of health professionals through AI models Wovenware worked on a proof of concept with a sample data set and our Innovation Sprint methodology. The team:
- Preprocessed data to remove constant variables and fill out missing values
- Performed an exploratory analysis to explore associations and correlations between variables
- Constructed three Random Forest Predictive Models
The proof of concept resulted in a validation of the feasibility of using machine learning models to aid health professionals in the decision-making process of selecting ulcer treatments. With 63%, 65% and 100% accuracy and p-values < 0.001, < 0.001, and 0.975 -0.2, all models may significantly improve with more sample data. Wovenware provided an assessment of the data requirements and next steps in order to build a more robust predictive model.