Electrical impedance tomography (EIT) is expected to be used in image reconstruction of accumulated visceral fat in the human abdomen, where the accumulated visceral fat may be related to a risk of metabolic syndrome. This work aimed to study the effects of the anisotropy of bioelectrical conductivity on image reconstruction by EIT of human abdominal cross-section. We developed a tomography device with 64 electrodes arranged in upper and lower rings surrounding the subject’s abdomen at the umbilical level. An alternating current of 1.0 mA_rms at a frequency of 500 kHz was applied through two adjacent electrodes in the upper ring, and the potential difference between every adjacent pair of electrodes in the lower ring was measured. From these voltage data, the cross-section of the abdominal conductivity was imaged by iterative calculations using the finite element method and nonlinear numeric optimization. When we considered the conductivity anisotropy in both transverse and longitudinal directions, the conductivity values obtained were within the physiologically plausible range reported in the literature. On the other hand, when the conductivity anisotropy was neglected, the central part of the image had extraordinarily low conductivity, and was outside the physiological range. The need to consider the anisotropic conductivity in image reconstruction was further confirmed by significantly reduced residual squared error in the optimization calculation.