We have recently demonstrated the imaging capabilities of a prototype 64-element 1 MHz concave array with 100 mm radius of curvature. This array was optimized for therapeutic applications using high-intensity focused ultrasound (HIFU). We have shown that this dual-mode ultrasound array (DMUA) has a therapeutic operating field (ThxOF) that extends by ±3 cm and ±2 cm around its geometric center in the axial and lateral directions, respectively. We have also shown that appropriate apodization and accounting for element directivity along with conventional synthetic aperture beamforming produce a 50 dB imaging field of view (IxFOV) larger than the ThxOF. In addition, the spatial registration of imaging targets is as accurate as commercially available scanners. In this paper, we present results from an image-based refocusing algorithm whereby images formed by the DMUA are used to identify a refocusing target and a set of critical points where the incident power is to be minimized. The algorithm is validated experimentally in tissue mimicking phantom with strongly scattering ribs placed between the DMUA and the target. These results demonstrate what is potentially the most powerful advantage of the use of DMUAs in image-guided surgery. Namely, the inherent registration between the imaging and therapeutic coordinate systems. This allows for direct definition of targets and any surrounding critical structures to be avoided to minimize the collateral damage. With these capabilities, DMUAs may provide a most powerful paradigm for image-guided surgery.