For Laparoscopy The use of 3D Segmentation and Classic 2D Optics

 

3D Segmentation A Semi-Automatic Process

The organ to segment is irregular the more it will be necessary to segment axial sections to obtain a precise and coherent result integrating all the variations of shapes. In general, segmentation performed almost on each cut for the uterus and every 3 to 4 cuts on the kidney. The model then checked, exported and “smoothed”. The software then used to create a usable 3D model from the 2D contouring .It is therefore a semi-automatic process. 

For a large number of reasons use of simulators is a prudent choice like Lap Trainer Box.

Three pitfalls are inherent in this method: it is operator dependent and therefore requires training. It is relatively long and thirdly it requires a sufficient definition of the imaging if not the 3D model too extrapolated and too far from reality especially if its volume is irregular and heterogeneous. In our work based on MRI, this segmentation carried out from raw DICOM images of the organ of interest. It requires sufficient resolution, i.e. thin sections every 2.5 mm. 

3D Model Recreate

In fact, the classic resolution of pelvic MRIs calls for cuts every 4 mm. Our goal was first recreate the 3D uterine model. A normal uterus is 30 to 40 mm wide; cuts every 4 mm cannot give an accurate rendering of the organ. In addition, some myomas are 10 to 15-mm. Their 3D model cannot created with only two to three MRI slices. For our second study model, the target organ measured on average 75 mm but above all the studied tumors measured 8 mm on average required an optimization of the MRI to have sections of 1.5 mm thick with a resolution of 0.4mm, much better than a conventional MRI. 

We initially used sequences. We now use high-resolution T2 sequences .This step therefore makes it possible to create a 3D model of the organ of interest, of the tumor to resected, but also of any anatomical element that could be relevant for the surgeon. The model thus created has a surface but no particular texture on its surface. We have segmented in our work the uterus, the kidneys pigs, tumors to resect and the uterine cavity. Other elements such as vascularization, excretory ducts, ureters, rectum, etc. may segmented.

Preoperative 3D Model

Preoperative 3D model corresponding to the segmentation of the uterus and myomas. On the left, the model shown in the MRI 3D model creation in intraoperative and initial registration.

The creation of the intraoperative 3D model will in fact allow the initial phase of registration), the second phase being the phase of follow-up and therefore of real-time registration developed in the next chapter. The objective of this intraoperative step is to produce a 3D model of the organ of interest. This 3D model made up of a set of geometric and photometric information. 

There are several techniques for reconstructing these 3D models from 2D digital images. We use the term Shape-from-X to group them together. X being the method and therefore the information used in the reconstruction. 

Schematically there are two categories. 

1. Photometric Techniques 

2. Geometric Techniques.

3D Reconstruction Robotic Surgery 

In fact, these methods often combined but most of the approaches have developed for robotic surgery with stereolaparoscopes and not monocular optics. The undeniable advantage of the robot is the systematic stereoscopic vision. It facilitates the 3D reconstruction and the registration in giving two images of the same object seen by two14 optics at a slightly different angle.

Several authors have reported the use of this stereoscopic vision in the creation of the intraoperative 3D model and in the registration. Despite these recent developments, 3D reconstruction methods are still unreliable and the authors suggest combining their technique with an optical or magnetic marker technique. Indeed, even with stereoscopy, it is necessary to be able to integrate the deformation of the tissues. 

The partial occlusion of the field due to the instruments, the variable reflectivity of the tissues. Stereoscopy therefore does not solve the entire problem, moreover it involves a very high cost in both hardware and software, and its availability in all operating theaters is far from certain.

Classic 2D Optics Laparoscopy 

“Classic” 2D optics are by far the most widely used in Laparoscopy because of their much lower cost than 3D optics, their ease of use and their performance . For more than 10 years, researchers have tried to find solutions to 3D reconstruction from a 2D image of a deformable object. The preoperative form of the uterus in particular is not the same as its intraoperative form. 

These changes are mainly due to cannulation and to the different external pressures applied to the uterus. Three teams are pioneers in this research on the reconstruction of deformable 3D objects.

Only our group works more specifically on medical applications. The technical solution developed by our team initially based on a 3D structure produced by shape-from-motion thanks to a point cloud on the surface of the uterine fundus or on the surface of the kidney. This allows a rigid reconstruction of the organ of interest. This system has already validated and has proven its robustness for many applications. 

Conclusion

For this, it is necessary to have a short video sequence of the organ of interest integrating both a movement specific to the organ and a movement of the camera. This sequence lasts from one minute to one and a half minutes.

For more information visit our website: www.gerati.com