What question did this study set out to answer?

This research aims to enhance medical captioning accuracy using advanced deep learning models, specifically Transfer Learning and Transformers.

February 22, 2026International Journal of Computational Intelligence Systems0 citationsOpen Access

Exploring the Synergy between Very Large Transformer and LSTM Models for Effective Medical Captioning from Videos to Text: The Impact of Captioning in Healthcare

AVAswiga R V MZMoin Ahmed ZahirOld Dominion University

Key Points

This research aims to enhance medical captioning accuracy using advanced deep learning models, specifically Transfer Learning and Transformers.
Comparative analysis of Transfer Learning through the MedVisionCapturer model and Transformers using CausalVLM.
Evaluation of model performance using BLEU, CIDEr, METEOR, and ROUGE-L scoring metrics.
Application of models on a limited dataset of CT scan recordings.
Transfer Learning model achieves a BLEU score of 83.34, CIDEr score of 89.23, METEOR score of 43.91, and ROUGE-L value of 73.41.
Transformer model scores: BLEU 71.42, CIDEr 74.20, METEOR 99.06, ROUGE-L 96.10, indicating competitive performance but lower overall efficacy in some metrics.

Abstract

In today’s rapidly evolving digital landscape, the demand for accurate and contextually relevant subtitles for image and video content, particularly in the medical domain, is increasingly critical. Despite the proliferation of visual data across various platforms, existing captioning systems often struggle due to variations in visual settings, complex temporal relationships, and nuanced semantics. Additionally, challenges such as limited datasets, privacy issues, and specialized annotation requirements make medical image captioning particularly difficult. To tackle these challenges, we investigate cutting-edge deep learning methodologies, specifically Transfer Learning and Transformer models, through a comparative analysis. Specifically, we focus on Transfer Learning through the MedVisionCapturer model and Transformer models using CausalVLM. Our findings reveal that the Transfer Learning model achieves notable performance with a BLEU score of 83.34, CIDEr Score of 89.23, METEOR Score of 43.91, and ROUGE-L value of 73.41 when tested on a limited set of CT scan recordings. In contrast, the Transformer model attains competitive yet lower scores: a BLEU score of 71.42, CIDEr Score of 74.20, METEOR Score of 99.06, and ROUGE-L value of 96.10. Therefore, this work underscores the promise of the advanced models introduced to improve the efficacy of automatic medical image captioning systems, ultimately fostering better health outcomes within an increasingly complex medical landscape.

اسأل الذكاء الاصطناعي

Bookmark

View Full Paper