Literature Review
Study antimicrobial resistance, wound infections, and risk factors in trauma care.
Proposal Focus
Research prototype
Clinical pressureSupport earlier awareness before full lab confirmation is available.
Risk stratificationCompare simple models using transparent evaluation metrics.
Responsible boundariesExplore decision support while keeping medical judgment central.
PPrecision
RRecall
FNFalse negatives
The Problem
The World Health Organization describes antimicrobial resistance as one of the top global public health and development threats. For instance, in 2019, bacterial antimicrobial resistance was responsible for over 1.27 million deaths and associated with over 4.95 million deaths worldwide. These statistics show that it is not a small problem; it is one of the main challenges facing medicine.
Why This Matters in My Community
In Ukraine, many people understand the immediate danger of conflict-related injuries, but fewer think about what happens in the days that follow. A 2025 study published in Communications Medicine found that multidrug resistance was present in 84.6% of bacterial isolates from wound samples in Ukrainian civilian hospitals. This project focuses on that less visible consequence: resistant infections spreading through systems that are already under pressure.
Can machine learning help identify patients at higher risk of multidrug-resistant wound infections in conflict-related trauma care before complete microbiological results are available?
My Research Approach
Conceptual Framework
Create a conceptual framework for early MDR wound infection risk stratification.
Python Prototype
If suitable public data is available, build a simple classifier and compare models such as logistic regression, decision trees, and random forests.
How I Would Evaluate It
I would evaluate the prototype using precision, recall, and false negatives. Recall and false negatives would be especially important because missing a high-risk case could be the most dangerous outcome.
Responsible Use and Limits
I would not make clinical conclusions independently, as I do not yet have a formal medical background. My strength would be in the technical side: building computational approaches, analyzing patterns, and learning responsible medical AI research under the guidance of a mentor in medical informatics or computational medicine.
Expected Deliverables
- Literature review
- Conceptual framework for early MDR wound infection risk stratification
- Python prototype, if data allows
- Model comparison and evaluation
- Clear limitations and ethical boundaries
Potential Impact
In overloaded wartime hospitals, earlier identification of patients at high risk for multidrug-resistant wound infections could mean faster laboratory prioritization, earlier infection control measures, and more informed clinical decisions before treatment options narrow. This project does not aim to build a clinical product. It aims to demonstrate that a research-based computational framework for this problem is feasible, responsible, and worth pursuing seriously. For me, that is the kind of computer science that matters: not tools built for ideal conditions, but tools designed for the moments when systems are under the most pressure and decisions cannot wait.