1
|
Numerical Evaluation on Parametric Choices Influencing Segmentation Results in Radiology Images—A Multi-Dataset Study. ELECTRONICS 2021. [DOI: 10.3390/electronics10040431] [Citation(s) in RCA: 4] [Impact Index Per Article: 1.3] [Reference Citation Analysis] [Abstract] [Track Full Text] [Subscribe] [Scholar Register] [Indexed: 01/17/2023]
Abstract
Medical image segmentation has gained greater attention over the past decade, especially in the field of image-guided surgery. Here, robust, accurate and fast segmentation tools are important for planning and navigation. In this work, we explore the Convolutional Neural Network (CNN) based approaches for multi-dataset segmentation from CT examinations. We hypothesize that selection of certain parameters in the network architecture design critically influence the segmentation results. We have employed two different CNN architectures, 3D-UNet and VGG-16, given that both networks are well accepted in the medical domain for segmentation tasks. In order to understand the efficiency of different parameter choices, we have adopted two different approaches. The first one combines different weight initialization schemes with different activation functions, whereas the second approach combines different weight initialization methods with a set of loss functions and optimizers. For evaluation, the 3D-UNet was trained with the Medical Segmentation Decathlon dataset and VGG-16 using LiTS data. The quality assessment done using eight quantitative metrics enhances the probability of using our proposed strategies for enhancing the segmentation results. Following a systematic approach in the evaluation of the results, we propose a few strategies that can be adopted for obtaining good segmentation results. Both of the architectures used in this work were selected on the basis of general acceptance in segmentation tasks for medical images based on their promising results compared to other state-of-the art networks. The highest Dice score obtained in 3D-UNet for the liver, pancreas and cardiac data was 0.897, 0.691 and 0.892. In the case of VGG-16, it was solely developed to work with liver data and delivered a Dice score of 0.921. From all the experiments conducted, we observed that two of the combinations with Xavier weight initialization (also known as Glorot), Adam optimiser, Cross Entropy loss (GloCEAdam) and LeCun weight initialization, cross entropy loss and Adam optimiser LecCEAdam worked best for most of the metrics in a 3D-UNet setting, while Xavier together with cross entropy loss and Tanh activation function (GloCEtanh) worked best for the VGG-16 network. Here, the parameter combinations are proposed on the basis of their contributions in obtaining optimal outcomes in segmentation evaluations. Moreover, we discuss that the preliminary evaluation results show that these parameters could later on be used for gaining more insights into model convergence and optimal solutions.The results from the quality assessment metrics and the statistical analysis validate our conclusions and we propose that the presented work can be used as a guide in choosing parameters for the best possible segmentation results for future works.
Collapse
|
2
|
Paolucci I, Sandu RM, Sahli L, Prevost GA, Storni F, Candinas D, Weber S, Lachenmayer A. Ultrasound Based Planning and Navigation for Non-Anatomical Liver Resections – An Ex-Vivo Study. IEEE OPEN JOURNAL OF ENGINEERING IN MEDICINE AND BIOLOGY 2020; 1:3-8. [PMID: 35402957 PMCID: PMC8979632 DOI: 10.1109/ojemb.2019.2961094] [Citation(s) in RCA: 2] [Impact Index Per Article: 0.5] [Reference Citation Analysis] [Abstract] [Grants] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 12/04/2019] [Revised: 12/16/2019] [Accepted: 12/17/2019] [Indexed: 01/10/2023] Open
Abstract
Goal: Non-anatomical resections of liver tumors can be very challenging as the surgeon cannot use anatomical landmarks on the liver surface or in the ultrasound image for guidance. This makes it difficult to achieve negative resection margins (R0) and still preserve as much healthy liver tissue as possible. Even though image-guided surgery systems have been introduced to overcome this challenge, they are still rarely used due to their inaccuracy, time-effort and complexity in usage and setup. Methods: We have developed a novel approach, which allows us to create an intra-operative resection plan using navigated ultrasound. First, the surface is scanned using a navigated ultrasound, followed by tumor segmentation on a midsection ultrasound image. Based on this information, the navigation system calculates an optimal resection strategy and displays it along with the tracked surgical instruments. In this study, this approach was evaluated by three experienced hepatobiliary surgeons on ex-vivo porcine models. Results: Using this technique, an R0 resection could be achieved in 22 out of 23 (95.7% R0 resection rate) cases with a median resection margin of 5.9 mm (IQR 3.5–7.7 mm). The resection margin between operators 1, 2 and 3 was 7.8 mm, 4.15 mm and 5.1 mm respectively (p = 0.054). Conclusions: This approach could represent a useful tool for intra-operative guidance in non-anatomical resection alongside conventional ultrasound guidance. However, instructions and training are essential especially if the operator has not used an image-guidance system before.
Collapse
Affiliation(s)
- Iwan Paolucci
- ARTORG Center for Biomedical Engineering ResearchUniversity of Bern Bern Switzerland
| | - Raluca-Maria Sandu
- ARTORG Center for Biomedical Engineering ResearchUniversity of Bern Bern Switzerland
| | - Luca Sahli
- ARTORG Center for Biomedical Engineering ResearchUniversity of Bern Bern Switzerland
| | - Gian Andrea Prevost
- Department of Visceral Surgery and Medicine, Inselspital, Bern University HospitalUniversity of Bern Bern Switzerland
| | - Federico Storni
- Department of Visceral Surgery and Medicine, Inselspital, Bern University HospitalUniversity of Bern Bern Switzerland
| | - Daniel Candinas
- Department of Visceral Surgery and Medicine, Inselspital, Bern University HospitalUniversity of Bern Bern Switzerland
| | - Stefan Weber
- ARTORG Center for Biomedical Engineering ResearchUniversity of Bern Bern Switzerland
| | - Anja Lachenmayer
- Department of Visceral Surgery and Medicine, Inselspital, Bern University HospitalUniversity of Bern Bern Switzerland
| |
Collapse
|
3
|
Eppenga R, Kuhlmann K, Ruers T, Nijkamp J. Accuracy assessment of target tracking using two 5-degrees-of-freedom wireless transponders. Int J Comput Assist Radiol Surg 2019; 15:369-377. [PMID: 31724113 PMCID: PMC6989619 DOI: 10.1007/s11548-019-02088-9] [Citation(s) in RCA: 3] [Impact Index Per Article: 0.6] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 06/19/2019] [Accepted: 11/04/2019] [Indexed: 12/22/2022]
Abstract
Purpose Surgical navigation systems are generally only applied for targets in rigid areas. For non-rigid areas, real-time tumor tracking can be included to compensate for anatomical changes. The only clinically cleared system using a wireless electromagnetic tracking technique is the Calypso® System (Varian Medical Systems Inc., USA), designed for radiotherapy. It is limited to tracking maximally three wireless 5-degrees-of-freedom (DOF) transponders, all used for tumor tracking. For surgical navigation, a surgical tool has to be tracked as well. In this study, we evaluated whether accurate 6DOF tumor tracking is possible using only two 5DOF transponders, leaving one transponder to track a tool. Methods Two methods were defined to derive 6DOF information out of two 5DOF transponders. The first method uses the vector information of both transponders (TTV), and the second method combines the vector information of one transponder with the distance vector between the transponders (OTV). The accuracy of tracking a rotating object was assessed for each method mimicking clinically relevant and worst-case configurations. Accuracy was compared to using all three transponders to derive 6DOF (Default method). An optical tracking system was used as a reference for accuracy. Results The TTV method performed best and was as accurate as the Default method for almost all transponder configurations (median errors < 0.5°, 95% confidence interval < 3°). Only when the angle between the transponders was less than 2°, the TTV method was inaccurate and the OTV method may be preferred. The accuracy of both methods was independent of the angle of rotation, and only the OTV method was sensitive to the plane of rotation. Conclusion These results indicate that accurate 6DOF tumor tracking is possible using only two 5DOF transponders. This encourages further development of a wireless EM surgical navigation approach using a readily available clinical system.
Collapse
Affiliation(s)
- Roeland Eppenga
- Department of Surgical Oncology, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX, Amsterdam, The Netherlands
| | - Koert Kuhlmann
- Department of Surgical Oncology, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX, Amsterdam, The Netherlands
| | - Theo Ruers
- Department of Surgical Oncology, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX, Amsterdam, The Netherlands.
- Nanobiophysics Group, Faculty TNW, University of Twente, Enschede, The Netherlands.
| | - Jasper Nijkamp
- Department of Surgical Oncology, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX, Amsterdam, The Netherlands
| |
Collapse
|
4
|
Eppenga R, Kuhlmann K, Ruers T, Nijkamp J. Accuracy assessment of wireless transponder tracking in the operating room environment. Int J Comput Assist Radiol Surg 2018; 13:1937-1948. [PMID: 30099659 DOI: 10.1007/s11548-018-1838-z] [Citation(s) in RCA: 4] [Impact Index Per Article: 0.7] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 01/29/2018] [Accepted: 07/27/2018] [Indexed: 01/23/2023]
Abstract
PURPOSE To evaluate the applicability of the Calypso® wireless transponder tracking system (Varian Medical Systems Inc., USA) for real-time tumor motion tracking during surgical procedures on tumors in non-rigid target areas. An accuracy assessment was performed for an extended electromagnetic field of view (FoV) of 27.5 × 27.5 × 22.5 cm (which included the standard FoV of 14 × 14 × 19 cm) in which 5DOF wireless Beacon® transponders can be tracked. METHODS Using a custom-made measurement setup, we assessed single transponder relative accuracy, absolute accuracy and jitter throughout the extended FoV at 1440 locations interspaced with 2.5 cm in each orthogonal direction. The NDI Polaris Spectra optical tracking system (OTS) was used as a reference. Measurements were taken in a room without surrounding distorting factors and repeated in an operating room (OR). In the OR, the influence of a carbon fiber and regular stainless steel OR tabletop was investigated. RESULTS The calibration of the OTS and transponder system resulted in an average root-mean-square error (RMSE) vector of 0.03 cm. For both the standard and extended FoV, all accuracy measures were dependent on transponder to tracking array (TA) distances and the absolute accuracy was also dependent on TA to OR tabletop distances. This latter influence was reproducible, and after calibrating this, the residual error was below 0.1 cm RMSE within the entire standard FoV. Within the extended FoV, this residual RMSE did not exceed 0.1 cm for transponder to TA distances up to 25 cm. CONCLUSION This study shows that transponder tracking is promising for accurate tumor tracking in the operating room. This applies when using the standard FoV, but also when using the extended FoV up to 25 cm above the TA, substantially increasing flexibility.
Collapse
Affiliation(s)
- Roeland Eppenga
- Department of Surgical Oncology, The Netherlands Cancer Institute, Amsterdam, The Netherlands
| | - Koert Kuhlmann
- Department of Surgical Oncology, The Netherlands Cancer Institute, Amsterdam, The Netherlands
| | - Theo Ruers
- Department of Surgical Oncology, The Netherlands Cancer Institute, Amsterdam, The Netherlands
- Nanobiophysics Group, MIRA Institute, University of Twente, Enschede, The Netherlands
| | - Jasper Nijkamp
- Department of Surgical Oncology, The Netherlands Cancer Institute, Amsterdam, The Netherlands.
- Department of Surgery, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX, Amsterdam, The Netherlands.
| |
Collapse
|
5
|
WEI XL, XUE MF, QIN ZX, BAI XY, DONG FF, ZHANG JJ, LV N, CHEN H, ZHANG J. The Effect of Comprehensive Care on the Patients Received Minimally Invasive Percutaneous Nephrolithotomy. IRANIAN JOURNAL OF PUBLIC HEALTH 2017; 46:923-929. [PMID: 28845403 PMCID: PMC5563874] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Download PDF] [Figures] [Subscribe] [Scholar Register] [Indexed: 11/21/2022]
Abstract
BACKGROUND We analyzed the effect of comprehensive care on the patients who received minimally invasive percutaneous nephrolithotomy (MPCNL). METHODS Patients hospitalized from 2013-2014 in Zhumadian Central Hospital (n=124) were enrolled and divided into two groups on random basis. The control group was treated with routine nursing model while the observation group was given comprehensive care additionally. The surgery time, degree of comfort, complications and successful cases, hospitalization time, sleep quality, nursing satisfaction and changes of systolic pressure, pulse and respiratory at different time were observed and analyzed. RESULTS The surgery time of the control group was significantly longer than that of observation group (P<0.05). The observation group felt more comfortable and showed more significant successful cases than the control group. Moreover, the hospitalization time were significantly reduced in observation group when compared with control group (P<0.05). The sleep quality of the observation group was significantly better than that of the control group (P<0.05). Before anesthesia, diastolic blood pressure, systolic blood pressure, pulse and respiration were not significantly different between the two groups. The diastolic blood pressure, systolic blood pressure, pulse and respiration after anesthesia, intraoperative 30 min, postoperative 30 min and other moments were significantly different. The incidence of complications in the control group was significantly higher than that in the observation group. The nursing satisfaction of the observation group was significantly higher than that of the control group. CONCLUSION The comprehensive care on the patients undergoing MPCNL was effective, and it can dramatically shorten surgery time, improve the success rate, improve the sleep quality of patients, keep life sign stable and minimize the complications.
Collapse
Affiliation(s)
- Xue-Li WEI
- Dept. of Urinary Surgery No.2, Zhumadian Central Hospital, Zhumadian, China,Corresponding Author:
| | - Mei-Fang XUE
- Physical Examination Center, Zhumadian Central Hospital, Zhumadian, China
| | - Zhao-Xia QIN
- Dept. of Urinary Surgery No.2, Zhumadian Central Hospital, Zhumadian, China
| | - Xing-Yun BAI
- Dept. of Urinary Surgery No.2, Zhumadian Central Hospital, Zhumadian, China
| | - Fang-Fang DONG
- Dept. of Urinary Surgery No.2, Zhumadian Central Hospital, Zhumadian, China
| | - Jin-Jin ZHANG
- Dept. of Urinary Surgery No.2, Zhumadian Central Hospital, Zhumadian, China
| | - Ning LV
- Dept. of Urinary Surgery No.2, Zhumadian Central Hospital, Zhumadian, China
| | - Hui CHEN
- Dept. of Urinary Surgery No.2, Zhumadian Central Hospital, Zhumadian, China
| | - Jia ZHANG
- Dept. of Urinary Surgery No.2, Zhumadian Central Hospital, Zhumadian, China
| |
Collapse
|
6
|
|
7
|
Zhang P, Liang Y, Chang S, Fan H. Kidney segmentation in CT sequences using graph cuts based active contours model and contextual continuity. Med Phys 2013; 40:081905. [DOI: 10.1118/1.4812428] [Citation(s) in RCA: 9] [Impact Index Per Article: 0.8] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/07/2022] Open
|