Deep Learning-Driven Speed-of-Sound Correction