Generating lymphoma ultrasound image description with transformer model

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Lymphoma is a prevalent hematologic tumor diagnosed via high-resolution ultrasound.
• Accurate characterization of diverse lymphoma ultrasonographic manifestations is challenging for sonographers.

Purpose of the Study:

• To develop a Transformer-based model for generating descriptive ultrasound images of lymphoma.
• To provide auxiliary guidance for ultrasound doctors during lymphoma screening.

Main Methods:

• A Transformer-based model incorporating deep stable learning to eliminate feature dependencies.
• Integration of a memory module in the decoder to enhance semantic information modeling and image depiction.

Main Results:

• The proposed model demonstrated superior prediction performance compared to existing methods.
• Experimental validation was conducted on an ultrasonic diagnosis dataset from Shanghai Ruijin Hospital.

Conclusions:

• The Transformer-based model offers improved diagnostic support for lymphoma using ultrasound.
• The model's advanced features enhance semantic understanding and detailed image generation for lymphoma ultrasonography.