Comparative studies of large language models versus human experts in supporting ultrasound in clinical practice: methodological challenges and interpretative limitations