Prediction of open urinary tract in laparoscopic partial nephrectomy by virtual resection plane visualization

Laparoscopic surgery has been preferred by many surgeons and patients because of shorter recovery time and less surgical invasion [1]. A partial nephrectomy shows treatment outcomes equal to those of a radical nephrectomy and preserves postoperative renal function [2‐4]. Therefore, laparoscopic partial nephrectomy has become an increasingly common practice for small renal tumors.

To achieve a safe procedure, preoperative information is especially important for laparoscopic surgery. Organs and large vessels around the operated site are difficult to recognize only with the limited two dimensional (2D) view of the laparoscope. It is important to preoperatively understand the anatomical features of the patient.

We developed the three dimensional (3D) image processing software VoTracer to preoperatively obtain significant information from computed tomography (CT) or magnetic resonance imaging (MRI) images. VoTracer supports traditional volume rendering; the user can observe 3D images by modifying local transparency and changing the viewpoint [5]. It also supports various image segmentation methods, such as thresholding [6], region growing [7], graph cut [8, 9], and contour-based segmentation [10, 11]; the user can quickly segment organ and tumor regions from 3D images with simple interaction.

One of the postoperative complications of partial nephrectomy is urine leakage. If the urinary tract is open during tumor resection, suturing the crack is required for the prevention of postoperative urine leakage. To formulate an operation strategy, it is useful to predict the margin of the urinary tract opening during the tumor resection.

In this study, we focused on laparoscopic partial nephrectomy and presented a method to predict the presence of an open urinary tract and the position of the opening using VoTracer. We also evaluated our method by comparing our predictions with actual operation results.

Table 1 summarizes the backgrounds of the patients, prediction results, and surgical outcomes. Table 1 also shows the minimum margins of the openings the evaluator detected in the virtual resection planes and actual surgical margins measured from the operation samples at the bottom of the tumor.In terms of presence of an open urinary tract (i.e. minimum opening margin), the predictions and intraoperative results approximately coincided in all patients. However, in patient No. 2, the predicted opening margin was 1 mm greater than the actual margin, and this may be explained by the size of the margin, which was likely to be larger at the site of urinary tract opening than at the bottom. The margin is generally the smallest at the bottom, and becomes larger at the periphery. We think this small difference was not critical for predicting the possible presence of an open urinary tract. The positions of the openings were similar to the predicted positions (Figure 3). In the two cases where the urinary tract was open, the patients were properly treated and no patient had postoperative complications. Surgical pathology reported clear cell renal cell carcinoma with negative margins in all cases.

In this study, we focused on open urinary tracts and developed a way to virtually visualize resection planes from CT images. Preoperatively understanding the 3D relationship of organs is particularly important for laparoscopic partial nephrectomy. In the past, surgeons had to infer the relationship of organs from CT slices; they performed 3D shape reconstructions in their heads. This required extensive experience and it was impossible to share ideas with others. In contrast, our method provides visualization that is close to the actual intraoperative view and allows surgeons to easily understand the 3D relationship of important organs. We believe that the preoperative information provided by our visualization method improves the safety of operations.

The major postoperative complications of laparoscopic partial nephrectomy are urine leakage, parenchymal bleeding and acute renal failure. Breda et al. reported that urine leakage of laparoscopic partial nephrectomy occurred in 1.4-2.0% of patients [13]. Thompson et al. reported that ischemia time is the most important factor in predicting renal function preservation [14]. To prevent complications, cracks should be securely sutured and ischemia time should be decreased, which can be accomplished by appropriate preoperative planning. For instance, the approach (trans-abdominal or trans-retroperitoneum), port position, and resection line should be selected depending on the tumor position and surrounding organs. Gill et al. reported on the novel technique of zero ischemia partial nephrectomy with preoperative planning using a visualized 3D view of the renal artery [15]. Our method virtually visualizes the intraoperative view from a different viewpoint. It may allow surgeons to visualize operation procedures in a different preoperative setting and enable them to select a better setting. Nephrometry scoring systems have been developed in an effort to standardize tumor assessment. R.E.N.A.L. Nephrometry Score (NS) has been proposed by Uzzo et al. [16]. Ficarra et al. introduced Preoperative Aspects and Dimensions Used for an Anatomical (PADUA) system. Our visualization technique contributes to more precise scoring of those nephrometry scores [17]. Our study focused on visualization of open urinary tracts, the first such study to have been reported.

Training for laparoscopic surgery is much more difficult than that for open surgery. Since laparoscopic surgery has a steep learning curve, it is important to provide sufficient training. Various laparoscopic surgery training systems have been developed. Some recent systems reproduces the entire procedure of laparoscopic radical nephrectomy [18, 19]. However, to the best of our knowledge, no other system provides an actual operative view of the individual patient. In contrast, our method provides visualization similar to the actual laparoscopic operative view using individual patient data.

The image segmentation software VoTracer, which we developed for this study, is useful in creating simulation models. Our research group has published studies on a mission rehearsal type surgical simulator that uses patient-specific models [20‐22]. In these studies, we reported that the patient-specific surgical simulator is useful for less-experienced doctors in practicing standard operations and for experts in developing new surgical procedures. However, one of the biggest obstacles of such a simulator is the process of creating simulation models from 3D images of patients. The image segmentation method presented in this study allowed us to segment each region (kidney, urinary tract, and tumor) in a few minutes. This drastically accelerates the model creation process. Combining the surgical simulator and the image segmentation method is an ongoing project that we hope to realize in the future.The major limitation of our current method is that the segmentation process requires user operation. The user has to set region growing parameters to correctly segment the urinary tract region. The user also has to specify contours to segment tumor regions based on his or her subjective knowledge (Figure 1D). Thus, segmentation results may vary depending on the user. Notice that this is retrospective study and the number of cases is also restricted. Larger population study is required in order to raise the accuracy of prediction. In the future, we would like to conduct a large-scale study and develop a segmentation method that minimizes user operation as much as possible. In this study, we used the delay phase of contrast enhanced CT in order to precisely segment the urinary tract. Due to CT imaging timing issues, we had several cases in which it was difficult to precisely segment the urinary tract. It is necessary to determine the best timing of CT imaging in order to stably segment urinary tract regions.

In this study, we developed a method to predict the presence of an open urinary tract and the position of the opening in laparoscopic partial nephrectomy from 3D CT images. The results revealed that we could make accurate predictions in 5 cases. This was a retrospective study and the number of cases was small. A larger scale study is required in order to confirm the accuracy of the predictions.