The new Leksell Gamma knife convolution algorithm requires evaluation prior to implementation in clinical practice. The superiority of this algorithm, which takes into account tissue electron densities, was evaluated using EBT3 GafChromicTM films within an anthropomorphic phantom. The CIRS anthropomorphic head phantom was chosen for its relevance to validate the convolution algorithm. Absolute dose and dose distributions were measured and compared with the outputs calculated from the Leksell Gamma Plan algorithms (TMR10 and the convolution algorithm). The measured absolute dose and the dose distributions in the homogeneous region of the anthropomorphic phantom were clearly in agreement with the dose distribution computed by the convolution algorithm. In a heterogeneous region where soft tissues contain a medium, such as aluminium, or an air gap, the measured dose profiles drastically changed, and only the convolution algorithm was able to correctly compute the dose to water in water. The convolution algorithm was able to take into account regions with high or very low electron densities such that the measured absolute dose was nearly equal to that computed by the convolution algorithm, with a common accepted dose measurement error of 2%.