This repository contains a research design proposal developed for Research Design and Applications for Data and Analysis. The project examines whether AI-assisted tools improve diagnostic accuracy and workflow efficiency in radiology, with a specific focus on CT abdomen interpretation across different radiologist experience levels.
Artificial intelligence is often presented as a transformative solution for radiology workflows, but existing evidence shows mixed results. Some studies report improvements in efficiency, while others find little or no benefit in diagnostic accuracy. This project is designed to move beyond broad claims and evaluate when, how, and for whom AI assistance is actually useful in clinical practice.
The proposed study uses a quantitative experimental design to compare performance across three conditions:
- Radiologist-only: Radiologists interpret CT abdomen scans without AI assistance
- AI-assisted: Radiologists interpret the same type of scans with AI-generated support
- AI-only: The AI model independently interprets the scans without human involvement
Healthcare organizations are being asked to invest in AI tools with the promise of faster and more accurate clinical decision-making. For hospital administrators, clinical leaders, and policymakers, it is important to understand whether these systems deliver measurable value in real workflows. This research is intended to provide an evidence-based framework for evaluating AI adoption in radiology rather than relying on hype or marketing claims.
This project was developed as part of coursework in Research Design and Applications for Data and Analysis. The goal is to propose a meaningful research question and design a realistic study capable of producing actionable insights.
|- The problem
|- Intended Audience
|- Existing Literature
|- Anticipated Impact
|- Main Research Question
|- Sub-Questions
|- Definitions
|- Experimental Treatment
|- Data
|- Sample
|- Hypotheses
|- Variables
|- Descriptive Statistics
|- Inferential Statistics