Tissue attenuation has shown potential for characterizing soft tissues for many years, but natural biological variability has limited its effectiveness in many applications as a sole indicator of tissue health. Recently however, there has been increased interest in measuring tissue attenuation due to the need to compensate for frequency-dependent attenuation when quantifying scatterer correlation length and scatterer concentration. Correlation length and concentration might be capable of distinguishing benign from malignant tumors if attenuation can be accurately estimated. In this study, the traditional attenuation-estimation algorithm based on measuring the down-shift in center frequency of the ultrasound backscattered signal with propagation depth was modified to correct for focusing along the beam axis. Whereas previous approaches required a reference phantom to correct for the focusing, which is more challenging at higher frequencies, this approach corrected for the focusing by assuming that the field pattern along the focal zone could be approximated by a Gaussian function. Based on this approximation, a correction term was introduced to compensate for the effects of focusing when estimating attenuation in the focal region. The algorithm was verified using computer simulations and an ex vivo tissue sample, both of which used a 33-MHz spherically focused transducer with a focal length of 9 mm and an f-number of 3. The algorithm was validated in computer simulations by moving the region of interest used to obtain the attenuation through the focal region. The algorithms' sensitivity to noise was also assessed by varying the frequency bandwidth used in the Gaussian fit to find the spectral-peak frequency from 15 to 50 MHz. The accuracy of the attenuation estimate in the computer simulations was on the order of 10% for all of the cases while the precision of the estimates varied from S to 35% depending on the available bandwidth. Similarly, the attenuation of the ex vivo tissue sample was 2.6±0.6 dB/cm-MHz using the developed algorithm compared to 2.5±0.4 dB/cm-MHz as measured using an insertion loss technique.