Optical techniques have been developed to acquire blood information (e.g., hematocrit [Hct], saturation of oxygen, thrombus) noninvasively and continuously in an artificial heart. For the practical use of an optical Hct measurement, Twersky's theory has been shown to be useful and have a good agreement in forward-scattered measurements. However, it was not applied to backward-scattered measurements, which can provide the measurement with a less demanding spatial requirement. Additionally, optimal measurement for accuracy is not well examined. Therefore, we developed an accurate Hct measurement in an artificial heart using current optical devices. To this end, we focused on optimizing an emitter-detector distance to provide a maximum optical path length. We attached optical emitter and detector fibers on Tygon tubing at various distances to measure forward- and backward-scattered light. Fresh bovine blood (Hct: 30-50%) was circulated in the tubing by a nonpulsatile artificial heart. We calculated the optical path length at various emitter-detector distances by fitting the measured optical outputs and the reference Hcts to Twersky's theory. Then, we performed Hct measurements. As a result, Twersky's theory is applicable not only to forward- but also to backward-scattered measurements in the physiogical Hct range. In both forward- and backward-scattered measurements, calculated optical path lengths become maximum at the same emitter-detector distance. The accuracy of Hct measurement is improved two to three times with the emitter-detector distance compared with other distances. The mean error is less than 1 Hct%. This result shows that an accurate Hct measurement is realized by selecting the optimal emitter-detector distance, which provides maximum optical path length defined by Twersky's theory. Our study provides a framework for the practical and less restrictive application of the optical Hct measurement to patients with an artificial heart.