Body surface detection method for photoacoustic image data using cloth-simulation technique

Photoacoustic tomography (PAT) is a novel modality that can visualize blood vessels without contrast agents. It clearly shows blood vessels near the body surface. However, these vessels obstruct the observation of deep blood vessels. As the existence range of each vessel is determined by the distance from the body surface, they can be separated if the position of the skin is known. However, skin tissue, which does not contain hemoglobin, does not appear in PAT results, therefore, manual estimation is required. As this task is very labor-intensive, its automation is highly desirable. Therefore, we developed a method to estimate the body surface using the cloth-simulation technique, which is a commonly used method to create computer graphics (CG) animations; however, it has not yet been employed for medical image processing. In cloth simulations, the virtual cloth is represented by a two-dimensional array of mass nodes. The nodes are connected with each other by springs. Once the cloth is released from a position away from the body, each node begins to move downwards under the effect of gravity, spring, and other forces; some of the nodes hit the superficial vessels and stop. The cloth position in the stationary state represents the body surface. The body surface estimation, which required approximately 1 h with the manual method, is automated and it takes only approximately 10 s with the proposed method. The proposed method could facilitate the practical use of PAT.