Research Deep Dive:RaDialog Copilot

AI-assistedradiologyreportingwithconfidence-basedtextsuggestionsandGitHubCopilot-styleinterface

Private academic project — no public demo or repository

👥 TUM × Johns Hopkins Collaboration🏆 Academic Research Project📅 2024

Research Overview

Confidence-aware copilot for radiology report drafting. Combines real-time feedback with calibrated suggestions: radiologists type findings, the system responds with confidence-weighted completions — speeding reporting while reducing AI bias.

Medical AI

Confidence Estimation

Model Calibration

Radiology Reporting

Human-in-the-Loop

Clinical NLP

Sentence Embeddings

Early Stopping

React

Flask

Entropy Analysis

Research Metrics

0+

X-ray Samples

0 Methods

Confidence Estimation

TUM + JHU

Academic Collaboration