Page 52 - Read Online
P. 52

Page 10 of 16            Kościuszko et al. Hepatoma Res 2021;7:51  https://dx.doi.org/10.20517/2394-5079.2021.17

               technique was selected. The paper presents values such as the time of surgery, bleeding and the need for
               blood transfusion concerning the tumour’s size. There were no postoperative complications in any of the
               patients. Patients were followed-up for one year - no recurrence occurred. The intraoperative appearance of
               the liver was consistent with the reconstruction image. Reduced blood loss and shorter surgery time were
               also found in reconstructive patients. It was also noted that such reconstructions can be useful for learning
               by students and young surgeons.

               Another experience was published by Wang et al.  Thirty patients requiring living donor liver
                                                              [60]
               transplantation were enrolled in the study. In this study, a preoperative simulation was carried out, and
               models were created on its basis. The segmentation method was not given-the organ volume was calculated
               automatically. The median age of recipients was 23 months in the study group and 11 months in the control
               group. The median age of the donors was 34 and 35 years, respectively. The authors pointed out the
               disadvantage of 3D reconstruction displayed on a flat screen. Printed models allow better 3D visualisation.
               Additionally, thanks to transparent silicone for parenchyma, it is easy to visualise the internal course of the
               vessels and the tumour’s location. Ultimately, no statistically significant difference was found for surgery,
               blood loss, length of stay, cost or survival. Fewer patients had complications, but the result was also
               statistically insignificant. The duration of donor surgery was significantly shortened. Follow-up lasted an
               average of 35 months.

               Esaki et al.  published a paper that is unlike the other works presented in this review. They proposed a
                        [61]
               method that uses a convolutional neural network (U-net), developed for biomedical image segmentation at
                                                                         [62]
               the Computer Science Department of the University of Freiburg . U-net is provided as open-source
               software. The method was used for automatic segmentation of liver graft volume of paediatric liver
               transplant patients. The accuracy of automatic segmentation was assessed after the network learned using
               100 patient datasets. The accuracy was measured using Dice similarity coefficients on 20 patient datasets.
               The accuracy rate for liver graft was 87.10% ± 4.70%. The authors used a neural network to accelerate the
               segmentation process. Computed tomography datasets were used for training of the neural network and for
               verification of its efficacy. The segmentations created by the U-net were compared to those made by an
               experienced radiology technician and verified by radiologists.


                       [63]
               Ishii et al.  described a single case of using a 3D-printed model for a 22-month-old girl with biliary atresia,
               leading to a liver transplant. An additional anatomical difficulty was situs inversus - a rare congenital defect
               causing all internal organs’ vertical inversion. During the model’s standard creation, the accuracy was
               assessed by scanning with an industrial CT system. After sterilisation, the printed model was taken to the
               operating room to compare its anatomy with the actual patient. The authors pointed out that the perception
               of the dorsal side of liver anatomy is more straightforward, thanks to the 3D model that could be easily
               rotated in the hands. The operation time was 646 min, blood loss was 425 mL, and the operation was
               completed without any intraoperative complications. The developers noted and quoted the time it takes to
               create the model: it took about 30-60 min to generate STL files from the original CT data. Then, it usually
               took about three days to produce 3D liver models from STL files. An additional day was needed for ethylene
               oxide sterilisation. Accordingly, the entire process takes 4-5 days.


               DISCUSSION (PERFECT)
               Complete surgical resection of hepatoblastoma is considered the most crucial step to achieve long-term
               survival . Open surgery is most commonly performed , but laparoscopic liver surgery is also a feasible,
                      [64]
                                                              [65]
               yet very rarely used, approach . The International Laparoscopic Liver Society members encourage routine
                                        [66]
               use of laparoscopy in local liver resections and left lateral segmentectomies . One of the reviewed studies,
                                                                               [67]
   47   48   49   50   51   52   53   54   55   56   57