Optimizing a Multimodal Large Language Model for Ultrasound-Based Thyroid Nodule Malignancy Classification: A Comparative Study of Few-Shot Learning, Prompt Engineering, and Fine-Tuning.

For GPT-4o, the hybrid strategy achieved the highest area under the receiver operating characteristic curve (AUC) in DDTI (0.866), TN5000 (0.689), and the atypia of undetermined significance cohort (0.836).

However, the hybrid model misclassified 12 of 43 malignant nodules, corresponding to a false-negative rate of 27.9%.

For Gemini 2.5 Flash-Lite, few-shot learning, fine-tuning, and the hybrid strategy did not improve AUC relative to text prompting in any dataset.