The Research Of 3D Classification Of Right Portal Vein And Computer-assissted Liver Segmentation Based On Portal Territory

Research BackgroundHepatocellular carcinoma is the sixth most commonly occurring cancer in the world and the second contributor to cancer mortality. Around 383000 people died annually in China, accounting for 51% of the world’s percentage. Liver resection and liver transplantation are the most promising methods to cure it or improve the survive rate. The principles of liver resection including complete excision of all tumors, protect remaining functional liver and decrease operative injury. Tumor invades into portal venous branches and tumor cells are carried to the distal part of the liver segments by portal venous flow and grow into daughter nodules is the main method of intrahepatic metastasis in HCC patients. Anatomical liver resection is the most preferred surgical procedure with complete excision the tumors and tumor-bearing portal territories that induce the risk of growing and recurrence of microscopic tumor thrombi. However, to locate the tumor and define the bearing portal venous branches is the premise of anatomical liver resection.Morphological observation portal vein mainly focus on the trunk of portal vein in previous of studies that usually give priority to large hepatic resection like sectionectomy or hemihepatectomy. And the most widely used is according to Cheng. The travel, distribution of portal vein is different from living state due to loss of support of blood pressure with the method of liver corrosion cast specimen. Besides, Terms describe the location of segment is relative to the desktop in vitro that the relationship between each segment was defined up and down which in contrast with the real dorsal and ventral relationship of liver segments inside abdomen.The removal unit of anatomical liver resection was segment or subsegment. The portal venous branch is punctured to stain the tumor bearing hepatic parenchyma and subsequently divided along the stained boundary so that three and four order portal vein need to be accurately observed. On the other hand, to avoid the transection plane from deviation, hepatic veins are used to guide the route. Therefore, it is essential to stereo observation of relationship of portal vein and hepatic vein.With the development of computer graphics technique and combination with medical imaging technology, three-dimensional reconstruction software realized the stereo visualization observation of live hepatic vessels. Large sample size is benefit of obtaining the real laws of travel and distribution of portal vein. The travel of left portal vein is simple:the transverse portion turning to corner portion and then transited to umbilical portion. P2 and P3 give off the umbilical portion towards the left and P4 towards the right. With the deep and tumor preferred location and without distinct anatomic landmark, the aberrant right portal vein is common and always the key and difficult area to anatomist and hepatic-biliary expert.Hence, the first part of this subject is to study the distribution laws and classify the 3D types of right portal vein by using 3D visualization software that jointly developed with yorktal digital medical imaging technology (shenzhen) co. LTD to provide morphological basis for anatomical right liver resection.Liver segmentation is the most effective method to locate the tumors. Namely, the eight segments scheme credited to Couinaud, has gained worldwide acceptance. Nevertheless, there sometimes may not be the case if aberrant portal venous appeared. Hepatic veins and portal system are used to divide Couinaud’s segmentation of the liver. However, The territory of each divided segments is not the basic unit of portal venous supplying. Anatomical liver resection demand complete excision of all tumors and the territories of tumor-bearing portal pedicles to avoid tiny PVTT residual lead to couinaud’s liver segment classification is not suitable for anatomic liver resection. One portal branch territory is classified as one resection unit by liver segmentation based on portal territory that in accord with the mechanism of anatomical liver resection that assist performing anatomical liver resection accurately and accepted by all of world’s hepatic-biliary experts.Liver segmentation based on portal territory firstly proposed by Japanese scholar, Takayssure, who think that the relationship between each other was dorsal and ventral. Despite the different volumes of two lobes, the right and left portal vein can be regarded as bilateral symmetry so that Cho divided the liver into 7 segments:caudate lobe, latero-superior segment, latero-inferior segment, medial segment, antero-ventral segment, antero-dorsal segment and posterior segment. Fasel proposed the concept of "1-2-20" by observing and counting numbers of third order portal branches and found that average 20(9-44) third order portal vein branches per people. He thought that the numbers of segments are not in contradiction with couinaud scheme and can be simplified to 8 when required.To determine the boundary and region of segment is by puncturing the target portal branches with the guide of IOUS and stain the liver parenchyma. However, the method only can be performed in the operation. How to predict the boundary and region of segment before surgery had troubled every hepato-biliary surgeons. With the development of 3D reconstruction software and many mathematical methods be proposed like LASA and NNSA, some software can determine the liver vessel territory and calculate the volume precisely that solve the problem before anatomical liver resection.Therefore, the second part of subject aimed to divide the portal vein belonging of each segment and guide the liver segmentation with the combination of yorktal digital medical imaging technology (shenzhen) co. LTD that founded by science and technology project of Guangzhou (21300000185) based on the 3D portal classification in the first part. To explore a computer-assissted liver segmentation method and examine the accuracy by using correlation and Bland-Altman method to analysis the consistency the simulated volume before and actual volume after surgery.Part Ⅰ.3D classification of the right portal vein based on three dimensional visualization technologyObjectives1. To investigate the effect of 3D classification of portal vein to anatomical liver resection.2. To build the 3D classification of right portal vein according to the anatomy study with three dimensional visualization technique.Materials and Methods1. SubjectsAbdominal CT enhancement data of 336 patients from March 2015 to June 2015 were collected.120 patients were enrolled in this study in accordance with specified criteria. There are 64 male and 56 female patients with mean age of 47.5 years (range from 10 to 85).2. Equipment and facilities(1) CT scanner:Obtain contrast enhanced multiphase CT scans using a 64-row or 256-slice CT scanner (CT PHILIPS Brilliance; The Netherlands); (2) Images were processed by a Mxview workstation; (3) HP ProLiant BL from digital center of southern medical university; (4) High-performance computer; (5) Three dimensional reconstruction software which co-developing with Xu Dong group(MI-3DVS, Medical Image Three-Dimensional Visualization System, granted number: 2008SR18798)3. The collection and storage of CT imagesThe scanning parameters and methods see the following reference:Fang C, Liu J, Fan Y, et al. Outcomes of hepatectomy for hepatolithiasis based on 3-dimensional reconstruction technique[J]. Journal of the American College of Surgeons,2013, 217(2):280-2884. Three-dimensional reconstructionImport multiphase DICOM images into MI-3DVS, the liver was segmented by using geodesic active contour model based on boundary information and portal vein and hepatic vein by using the modified Hessian matrix of vascular enhancement algorithm and region-growing algorithm. Marching Cubes algorithm is used to reconstruction the segmentation data. The travel, distribution and variation of the branches of right portal vein were observed.5. Individualized computer-assisted liver segmentationExtraction portal centerline using territory analysis module based on the third-order portal branches and it’s boundary of watershed. The surface boundaries of the liver segments and intersegmental plan are divided using the nearest neighbor segment approximation algorithm.Results1. Three dimensional reconstruction resultsDistribution and travel of the third-and fourth-order portal branches and relation with hepatic veins are clearly displayed. Individualized anatomy of portal and hepatic vein are truly reflected.2. Classification of main portal veinClassification criterion of main portal vein is referred to Cheng and Atri. The right (RPV) and left portal veins (LPV) were presented as independent trunks and the RPV were divided into two branches supplying the right anterior and right posterior segment of the liver in 102 cases (85.0%; Type I).In 9 cases (7.5%), the right anterior portal vein(RAPV) and right posterior portal vein(RPPV),and LPV were presented as the trifurcating of the main portal vein(MPV; Type II). In 5 cases (4.2%), the RPPV came from the MPV directly, and the LPV and RAPV were presented as common trunk (Type Ⅲ).In 4 cases (3.3%), the RPPV came from the MPV directly and the RAPV came from the LPV (Type IV).3. Relation of RPPV with the RHV and MHVDorsal branch of P8 reach over the trunk of the right hepatic vein (RHV) in 102 cases (85.0%) and 18 cases had not. Due to presence of large inferior right hepatic vein (IRHV) with short of RHV that branch of P8 could not reach over in 8 cases (6.7%). Ventral branch of P8 reach over the trunk of the middle hepatic vein (MHV) in 68 cases (56.7%).4. Classification of the right portal vein3D branching types of P5, P6, P7, P8 can be classified according to the branching points, travel, direction of their third-order branches. P5 was classified into five types:P5 branched from the trunk of the right paramedian sector (RPMS) in 31 cases (Type A,25.8%). P5 branched from the dorsal and ventral branches of P8 in 10 cases (Type B,8.3%). P5 branched from the trunk of the RPMS and branches of P8 in 43 cases (Type C,35.8%). P5 directly branched from the trunk of the RPMS in 7 cases (Type D,5.8%). P5 branched from the trunk of the right lateral sector (RLS) in 29 cases (Type E,24.2%). P8 was classified into four types:P8 included dorsal branch and ventral branch in 47 cases (type A; 39.2%). P8 included dorsal branch, ventral branch and lateral branch in 39 cases (type B; 32.5%). P8 included dorsal branch, ventral branch and medial branch in 16 cases (type C; 13.3%). P8 included dorsal branch, ventral branch, medial branch and lateral branch in 18 cases (type D; 15.0%). P6 was classified into four types:P6 branched from the trunk of the right lateral sector (RLS) in 51 cases (Type A,42.5%).P7 and P6 diverged from the common trunk. P6 branched from the trunk of the right portal vein in 18 cases (Type B,15.0%). P6 were the caudal branches of arching right posterior portal vein, in 39.2% of patients (type C; 47).P6 partly branched from the right paramedian sector (RPMS) in 4 cases (Type D,3.3%). P7 was classified into six types:P7 branched from the trunk of the right lateral sector (RLS) in 41 cases (Type A,34.2%).P7 and P6 diverged from the common trunk. P7 branched from the trunk of the right portal vein in 17 cases (Type B,14.2%). P7 were the headward branches of arching right posterior portal vein, in 33.3% of patients (type C; 40). Anther P7 branched from the trunk of the right posterior portal vein in 7 cases (Type D; 5.8%); Anther P7 branched from P7 in 12 cases (type E; 10.8%); Anther P7 branched from the crotch of the left and right portal vein in 2 cases (type F; 1.6%).5. Individual liver segmentationThe boundary of the liver surface and intersegmental plan of segments were divided according to third-order portal vein. Relation of vasculars can be clearly observed by hyalinizing the liver segments.Conclusions1. Three dimensional visualization technique is benefit of summarizing the laws that loyal to real anatomical portal vein structure as the complicated travel and lots of aberrance with right portal vein.2.3D classification of right portal vein provide the morphological basis for anatomical right liver resection and is beneficial to accurately perform anatomical right liver resection.Part Ⅱ. The research of individualized computer-assissted liver segmentation based on portal territoryObjectives1. To investigate a computer-assissted liver segmentation method based on portal territory.2. To investigate the value of the liver segmentation method in anatomical liver resection.Materials and Methods1. Subjects65 patients with HCC were performed liver resections at zhujiang hospital, southern medial university from December 2014 to February 2016.16 of them were performed anatomical liver resection based on operation planning with MI-3DVS.14 cases were man and two cases were female patients. The average ages were 47.7±10.1 years. The average tumor size was 5.27±2.64 cm and two patients were suffered repeat hepatectomy. Tumor size about 5cm was found in 9 patients and in 7 less than 5cm. The hepatitis B virus was infected in 12 patients,1 had hepatitis C infection and 1 both had hepatic B and hepatitis C infection.14 patients were diagnosed with cirrhosis according to image.2. Equipment and facilities(1) PHILIPS BRILLIANCE 64/256 rows of spiral CT that come from The Dutch Philip company; (2) High-pressure injection pump comes from Medrad stellant company; (3) Non-ionic contrast injection, ultravist; (4) Image post-processing workstation:Mxview workstation (CT own system) that comes from The Dutch Philip company.(5) Digital medical clinical trial center, southern medical university, HP blade server; (6) Computer system:Intel Core i5,4 nuclear,64bit CPU,8G RAM, 20G hard-disk space,64 bit Windows 7 operating system, nVidia, graphics card(1G video memory),1280*1024 displayer;(7)MI-3DVS (The Medical Image Three-Dimensional Visualization System, grant number:2008SR18798) joint developed with yorktal digital medical imaging technology (shenzhen) co. LTD.3. The collection and storage of CT imagesAs mentioned in the first part.4. Segmentation of liver and portal veinAs mentioned in the first part.5. Computer-assissted liver segmentation processComputer-assisted liver segmentation based on portal territory including three aspects:centerline extraction; mark the liver segments based on the 3D classification of portal vein in the first part; Divide the segment according to the marked centerline.5.1 Extraction portal vein centerlineThe segmented portal vein model was refined into centerline according to iteration thinning method. The simplified process of centerline cannot change topology relation of the original vascular tree that by refining voxel of vessel surface until not to be refined.5.2 Mark portal vein centerlineThe portal tree model was stored into tree data structure and marked the nodes on the tree data structure to correspond to different segments according to the distribution of portal vein distribution.5.3 Divided liver segments according to the marked portal vein centerlineTo divide the intersegmental demarcation and plane corresponding to the third order portal branches according to the Nearest Neighbor Segment Approximation (NNSA) method.6. Operation procedureThe right subcostal incision or Inverted "Y" shaped incision were chose according to specific situation. The first step was to dissect the hepatic portal and find the portal trunk of liver lobes after a laparotomy. To right lateral sectoriectomy or right paramedian sectoriectomy, indigo carnime dye was used to stain the parenchyma. 20G needle was used to puncture the portal branches with 4ml per segment. To add the dose and increase the injection speed depending on the stained situation. Tightening the occlusion string and marking the stained surface of the liver. Occlusion the right anterior portal vein or right posterior portal vein temporarily to mark the demarcating line and incision the part of liver parenchyma from the cephalic to the caudal portion. The landmark portal vein branches can be found and ligated after deepening the dissection depth to clear the boundary of segments.7. Statistical analysisTo calculate the simulate volume of tumor bearing liver segment before surgery and actual volume of liver segment after surgery by using dewatering method. Pearson correlation and Bland-Altman method were used to evaluate the consistency of the simulate volume of tumor bearing liver segment before surgery and actual volume of liver segment.Results1. The intersegmental boundary and plane can be observed clearly based on the 3D liver segment model. The interflow and outflow can be clearly observed by hyalinizing the liver segments.2. The intersegmental boundary and plane simulated before surgery have a good consistency with the actual surgery.3. The simulate volume of tumor bearing liver segment before surgery based on MI-3DVS and actual volume of liver segment after surgery have a high correlation.Equation:y=1.023x-17.716, R2=0.914. The mean difference of volume was-8.06ml and the standard deviation was 49.33ml with Bland-Altman method. The volume before and after surgery have a good consistency with 95% consistency limit-8.06±1.96×49.33,namely(-104.75,88.63).Conclusions1. The computer-assisted liver segmentation based on three order portal branch territory can real display anatomical structure of each segment and provide 3D anatomical and functional information.2. The computer-assisted individual liver segmentation based on portal territory is beneficial to anatomical liver resection.