Transcranial sonography (TCS) of small deep brain structures, such as substantia nigra and brainstem raphe, is increasingly used for assessment of neurodegenerative disorders. Still, there are reservations against TCS because of the smallness of evaluated structures and constraints on image resolution that is discussed to be lower compared to magnetic resonance imaging (MRI). To evaluate two different-generation TCS systems in visualizing fine intracranial structures, we studied image resolution on a phantom consisting of 0.80 mm x 1.05 mm regular meshwork of nylon threads embedded in a wet, gel-filled ex vivo human skull. Imaging was performed with a former-generation and a present-day clinical ultrasound system and for comparison with MRI. In axial direction of insonation both TCS systems resolved 0.80-mm and 1.05-mm thread-to-thread distance at depths between 55 and 120 mm using transmission frequencies > or =2.5 MHz. The meshwork, however, was recognizable as such only with the contemporary TCS system at depths between 60 and 85 mm due to its higher lateral resolution. MRI resolved the meshwork if image resolution was chosen sufficiently high but not if realistic clinical conditions were applied with its trade-offs between image SNR, resolution, total scan time, and unavoidable head motion during the latter. Hence, if the requirements for optimal TCS image resolution are fulfilled, i.e. sufficient acoustic bone window, increased echogenicity of target structure and its localization in a distance of maximum +/-15 mm from midsagittal plane, findings suggest that contemporary TCS systems achieve higher image resolution of intracranial structures in comparison not only to former-generation systems, but also to MRI under clinical conditions.