Transcranial MR-guided focused ultrasound (tcMRgFUS) is non-invasive treatment for neurological disorders such as essential tremor. However, due to RF field reflection from the surface of the transducer, the ultrasound transducer causes a curved dark band in midbrain regions, degrading image quality. Previous Studies have proposed passive antennas and thin copper screen to mitigate this artifact. This study proposes a new approach to reduce the dark band artifact using a deep learning based Residual U-Net that allows preserving brain structures in tcMRgFUS images.