Title:	HUMAN-IN-THE-LOOP AI FOR RADIOLOGY REPORT GENERATION: A METHODOLOGY FRAMEWORK FOR CONTINUOUS LEARNING FROM EXPERT FEEDBACK
Author(s):	Dharmik Savani and Firmino Oliveira da Silva
ISBN:	978-989-8704-71-9
Editors:	Paula Miranda and Pedro Isaías
Year:	2025
Edition:	Single
Keywords:	Artificial Intelligence, Human-in-the-Loop, Radiology, Report Generation, Medical Imaging, Natural Language Processing
Type:	Full Paper
First Page:	44
Last Page:	51
Language:	English
Cover:
Full Contents:	if you are a member please login
Paper Abstract:	The integration of artificial intelligence into radiological workflows faces challenges in clinical acceptance and diagnostic reliability. While automated report generation systems show technical feasibility, deployment is limited by concerns over accuracy, transparency, and alignment with clinical expertise. Current systems often operate as static models that cannot adapt to institutional preferences or learn from expert corrections, leaving a gap between algorithmic capability and clinical needs. This study introduces a human-in-the-loop methodology framework that combines computer vision analysis of medical images with natural language processing for report generation, while systematically capturing radiologist feedback to enable continuous model improvement. The architecture includes four modules: CNN-based image feature extraction, fine-tuned biomedical language models for report generation, a human interface for capturing radiologist feedback, and a feedback mechanism for continuous system improvement. By maintaining human oversight while leveraging machine-learning capabilities, the framework addresses key challenges in medical AI deployment, enhancing workflow efficiency and diagnostic consistency. Iterative feedback integration enables alignment with institutional reporting standards and adaptation to local clinical practices, fostering personalized AI assistance. The main objective is to design a human-in-the-loop AI framework for radiology that adapts to institutional practices and radiologist feedback, improving both workflow efficiency and diagnostic consistency. The methodology incorporates comprehensive evaluation strategies, assessing both technical performance and clinical utility to ensure that improvements translate into meaningful workflow gains. This approach advances the development of collaborative human-AI systems in healthcare, offering a structured path toward radiology report automation that prioritizes technological innovation, clinical utility, and the complementary role of AI.
	Go Back