Tractor drivers’ health is so important and some injuries can develop in the long run such as lower back pain or spinal column disorder resulting from low-frequency vibrations. Seat suspension plays a crucial role in the reduction of harmful vibrations in off-road vehicles. Passive suspension design parameters should be tuned properly to remove vibrations efficiently. Lumped parameter models have been widely used in seat suspension optimization to simulate human body responses. In this study, a tractor seat suspension was designed based on an artificial neural network biodynamic model instead of a lumped model. In the optimisation step, a novel approximation approach via an artificial neural network was employed to minimise seat-to-spine vibration transmissibility. Then, a seat suspension was modified based on the optimisation results, and an outdoor dynamic test was carried out to evaluate suspension in vibration transmissibility attenuation.  The gathered data were processed and analyzed in terms of vibration transmissibility, mobility, and apparent mass. The transmitted vibration from floor to spine was measured between 0.2 and 0.4 at 4 to 6 Hz (spine dominant frequency range) on bumpy and smooth roads at various traveling speeds. The effect of seat suspension modification showed the changes in the performance of seat suspension in regard to vibration transmissibility. Thus, this passive suspension may decrease the risk of lower back pain for tractor drivers. So, this modification can be considered by the tractor manufacturer for future designed models to provide more comfort for operators in the daily use of this device.