This study presents a few-shot learning approach for classifying ultrasound breast cancer images using meta-learning algorithms. The authors used prototypical networks (ProtoNet) and model-agnostic meta-learning (MAML) to address the challenges of skewed class distribution and limited labeled data in medical datasets. The Breast Ultrasound Images (BUSI) dataset, which has three classes, was used for meta-testing in a cross-domain approach, leveraging other datasets for meta-training. The proposed method achieved an accuracy range of 0.882–0.889, representing a significant improvement over the baseline accuracy of 0.831. ProtoNet outperformed MAML in all k-shot settings, and using ResNet models as the backbone network for feature extraction was more effective than a four-layer convolutional model. This study is the first to apply meta-learning for few-shot classification on the BUSI dataset, providing a promising solution for breast cancer diagnosis with limited data. The methodology can be adapted to other datasets with similar challenges.This study presents a few-shot learning approach for classifying ultrasound breast cancer images using meta-learning algorithms. The authors used prototypical networks (ProtoNet) and model-agnostic meta-learning (MAML) to address the challenges of skewed class distribution and limited labeled data in medical datasets. The Breast Ultrasound Images (BUSI) dataset, which has three classes, was used for meta-testing in a cross-domain approach, leveraging other datasets for meta-training. The proposed method achieved an accuracy range of 0.882–0.889, representing a significant improvement over the baseline accuracy of 0.831. ProtoNet outperformed MAML in all k-shot settings, and using ResNet models as the backbone network for feature extraction was more effective than a four-layer convolutional model. This study is the first to apply meta-learning for few-shot classification on the BUSI dataset, providing a promising solution for breast cancer diagnosis with limited data. The methodology can be adapted to other datasets with similar challenges.