Abstract
Ultrasound imaging, known for its non-invasive and real-time capabilities, has become an indispensable tool in the rapid diagnosis of numerous medical conditions. Its ability to provide immediate insights without the need for surgical intervention has revolutionized modern healthcare. With the advent and widespread adoption of portable and handheld ultrasound devices, ultrasound scanning has become more accessible to a larger population. While handheld ultrasound devices have made imaging more accessible, they often exhibit significant differences in resolution compared to other types of ultrasound devices. Differences in other technical features, such as penetration depth and frame rate, alongside resolution disparities, can also pose significant challenges, if not properly handled. To develop an AI solution that performs reliably across various ultrasound device models, it is crucial to ensure that the training data is not biased toward specific device types. Proper labeling of ultrasound device types within the dataset is essential to address this challenge. This research project aims to develop a semi-supervised approach for classifying the types of ultrasound devices from ultrasound images.