In recent years, natural orifice transluminal endoscopic surgery (NOTES) using a flexible endoscope is focused on to reduce burden of patients. Since NOTES is difficult for surgeons, flexible endoscopic robots also have been studied and developed. However, conventional robots for NOTES cannot transmit force sensation adequately. The lack of force sensation causes undesired injury. In this paper, a novel haptic forceps robot for NOTES is developed, and force sensation is transmitted by implementing four-channel bilateral control.

This paper proposes a structure-by-structure recognition method of spinal columns, ribs, intervertebral disks and vertebrae in abdominal X-ray CT images. First, bone regions are extracted by applying binarization operators to CT images, and then spinal columns and ribs are extracted from the bone regions by using morphological operators. Intervertebral disks are emphasized, and then plane models are fitted to the intervertebral disks to recognize their positions and directions. The spinal columns are divided into vertebrae based on the plane models. Spinal processes have more irregular shapes. In order to recognize vertebrae including the spinal processes, the plane models are converted into curved surface models that are represented by control points. The curved surface models are transformed by minimizing an energy function to evaluate (1) the smoothness of the curved surface models and (2) the fidelity of the model to pixel-value distributions in the images. By using the plane and curved surface models, thoracic and lumbar vertebrae are recognized. The proposed method was applied to fifteen cases and was able to correctly recognize 95% of ribs, 100% of intervertebral disks, and 75% of vertebrae. The experimental results demonstrated that the proposed method was promising as a means of recognizing bone regions in abdominal X-ray CT images.