Researchers have developed a method to adapt a 3D medical vision-language model (VLM) for institution-specific classification of hepatocellular carcinoma (HCC) using computed tomography (CT) scans and radiology reports. The study compared full fine-tuning and parameter-efficient fine-tuning (PEFT) techniques, finding that both approaches improved performance over the pretrained model without requiring lesion-level annotations. PEFT, specifically using ConvLoRA for the image encoder and LoRA for the text encoder, achieved comparable results to full fine-tuning while updating only about 1% of the model's parameters, though it required more training steps. AI
IMPACT Demonstrates efficient adaptation of medical VLMs for institution-specific tasks, potentially improving diagnostic accuracy with limited data.
RANK_REASON The cluster contains a research paper detailing a novel method for adapting a medical vision-language model. [lever_c_demoted from research: ic=1 ai=1.0]
- 3D Abdominal Vision-Language Model
- computed tomography
- ConvLoRA
- Febryan Putra Kartika
- Full Fine-tuning
- Hartford HealthCare
- LoRA+
- peft
- Radiology reports
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →