Purpose: To evaluate the utility of an ultrasoundscatterer size estimation method (combined with B‐mode features) to differentiate liver hemangioma (benign) from malignant liver tumors in a clinical study. Method and Materials:Scatterer size is estimated from the frequency dependence of the backscatter coefficient, obtained using a reference phantom method. A tissue mimicking phantom containing glass beads randomly distributed in gel was scanned by a Siemens Antares scanner equipped with a linear array transducer to verify the scatterer size data reduction methodology prior to use in clinical determinations. Excised normal human livertissues and in vivoliver hemangiomas and malignant tumors were evaluated using the methodology using either the Siemens Antares or GE Logic 9 scanners. The results were correlated with biopsy findings. Results: For power spectra estimation, Welch's average approach yields better results than a periodogram and an autoregressive approach. Averaging over 15 independent samples with a gating window length of around 10 wavelengths is desirable to obtain reasonable size estimates based on the trade off between spatial resolution and signal‐to‐noise ratio. Therefore, angular or elevational compounding was used for in vivo data acquisition. A liver hemanigomas exhibited larger effective scatterer sizes than surrounding normal livertissues. Combining B‐mode features, such as a halo sign, with scatterer size estimates may be useful to confirm the benign or malignant classification. Conclusion:Ultrasoundscatterer sizes estimated from frequency dependent backscatter properties may provide useful clinical information, not readily available in conventional ultrasound B‐mode images. Preliminary in vivo results suggest that it may be an effective processing method for distinguishing liver hemangiomas.