Related Experiment Video
Updated: Jul 7, 2026
08:32
Adapting Human Videofluoroscopic Swallow Study Methods to Detect and Characterize Dysphagia in Murine Disease Models
Published on: March 1, 2015
22.0K
A Vision-Language-Guided Multimodal Fusion Network for Glottic Carcinoma Early Diagnosis: Model Development and
Zhaohui Jin1, Yi Shuai2, Yun Li2
1College of Big Data and Internet, Shenzhen Technology University, Pingshan District, 3002 Lantian Road, Shenzhen, Guangdong, 518118, China, 86 19276679344.
JMIR Medical Informatics
|October 8, 2025
Summary
A new vision-language model, VLMF-Net, improves early diagnosis of glottic carcinoma (GC) by integrating text and images. This AI tool shows superior accuracy and robustness compared to existing methods, aiding early detection efforts.
Area of Science:
- Artificial Intelligence in Medicine
- Medical Imaging Analysis
- Natural Language Processing for Healthcare
Background:
- Early diagnosis of glottic carcinoma (GC) is crucial for patient prognosis but remains challenging due to visual similarities with benign conditions and limited access to expertise in underserved regions.
- Current diagnostic methods struggle with subtle morphological differences, necessitating advanced technological solutions for improved accuracy and accessibility.
Purpose of the Study:
- To develop and validate a novel vision-language multimodal model, VLMF-Net, for the efficient and accurate early diagnosis of glottic carcinoma.
- To overcome the limitations of existing technologies in detecting early-stage GC, particularly in resource-limited settings.
Main Methods:
- A vision-language-guided multimodal fusion network (VLMF-Net) was designed, integrating a Large Language Model (LLaMa) for text processing and a vision transformer for laryngoscopic image analysis.
- Cross-modal alignment was achieved using the Q-Former module, followed by feature fusion for deep integration of text and image data to enable automated classification diagnosis.
- Model performance was evaluated against baseline methods (CLIP, BLIP-2, ALIGN, VILT) using metrics including accuracy, recall, precision, F1-score, and AUC on both internal and external test sets.
Main Results:
- VLMF-Net achieved a 77.6% accuracy on the internal test set, outperforming the best baseline (BLIP-2 at 71.5%) by 6.1 percentage points.
- On the external test set, VLMF-Net demonstrated robust performance with 73.9% accuracy, surpassing the second-best model (BLIP-2 at 69.3%) by 4.6 percentage points.
- The results indicate VLMF-Net's superior diagnostic capability and strong generalization ability for early glottic carcinoma detection.
Conclusions:
- The VLMF-Net model offers an effective solution for the early diagnosis of glottic carcinoma.
- This AI-driven approach addresses key challenges in early GC detection, improving diagnostic efficiency and accuracy.
Keywords:
clinical decision makingcomputer-aided diagnosisglottic carcinoma early diagnosislarge-scale foundation modelmultimodal deep learningMore Related Videos
Related Concept Videos
02:43Mouse Models of Cancer Study
Mice have long served as models for studying human biology and pathology because of their phylogenetic and physiological similarity with humans. They are also easy to maintain and breed in the laboratory, and hence, many inbred strains are now available for research. Studies on mice have contributed immeasurably to our understanding of cancer biology.
The development of transgenic, knockout, and knock-in mice has led to an exponential increase in their use as model organisms in research,...
The development of transgenic, knockout, and knock-in mice has led to an exponential increase in their use as model organisms in research,...
02:43Mouse Models of Cancer Study
Mice have long served as models for studying human biology and pathology because of their phylogenetic and physiological similarity with humans. They are also easy to maintain and breed in the laboratory, and hence, many inbred strains are now available for research. Studies on mice have contributed immeasurably to our understanding of cancer biology.
The development of transgenic, knockout, and knock-in mice has led to an exponential increase in their use as model organisms in research,...
The development of transgenic, knockout, and knock-in mice has led to an exponential increase in their use as model organisms in research,...