Access the full text.
Sign up today, get DeepDyve free for 14 days.
Hindawi Applied Bionics and Biomechanics Volume 2019, Article ID 9756842, 14 pages https://doi.org/10.1155/2019/9756842 Research Article Real-Time Needle Force Modeling for VR-Based Renal Biopsy Training with Respiratory Motion Using Direct Clinical Data 1 1 1 2 1 1 Feiyan Li, Yonghang Tai , Qiong Li, Jun Peng, Xiaoqiao Huang, Zaiqing Chen, and Junsheng Shi Yunnan Key Laboratory of Opto-electronic Information Technology, Yunnan Normal University, Kunming, China Department of Urology Surgery, Yunnan First People’s Hospital, Kunming, China Correspondence should be addressed to Yonghang Tai; email@example.com Received 11 July 2018; Revised 26 March 2019; Accepted 14 May 2019; Published 25 June 2019 Academic Editor: Agnès Drochon Copyright © 2019 Feiyan Li et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Realistic tool-tissue interactive modeling has been recognized as an essential requirement in the training of virtual surgery. A virtual basic surgical training framework integrated with real-time force rendering has been recognized as one of the most immersive implementations in medical education. Yet, compared to the original intraoperative data, there has always been an argument that these data are represented by lower ﬁdelity in virtual surgical training. In this paper, a dynamic biomechanics experimental framework is designed to achieve a highly immersive haptic sensation during the biopsy therapy with human respiratory motion; it is the ﬁrst time to introduce the idea of periodic extension idea into the dynamic percutaneous force modeling. Clinical evaluation is conducted and performed in the Yunnan First People’s Hospital, which not only demonstrated a higher ﬁtting degree (AVG: 99.36%) with the intraoperation data than previous algorithms (AVG: 87.83%, 72.07%, and 66.70%) but also shows a universal ﬁtting range with multilayer tissue. 27 urologists comprising 18 novices and 9 professors were invited to the VR-based training evaluation based on the proposed haptic rendering solution. Subjective and objective results demonstrated higher performance than the existing benchmark training simulator. Combining these in a systematic approach, tuned with speciﬁc ﬁdelity requirements, haptically enabled medical simulation systems would be able to provide a more immersive and eﬀective training environment. 1. Introduction models, and cadavers for surgical training, but since the material characteristics of artiﬁcial models are quite diﬀerent Minimally invasive surgery (MIS) is prevalent in the medical from real human tissues, they cannot achieve the eﬀect of training because of the lack of reality. Apart from that, the ﬁeld due to the characteristics of eﬃciency, less bleeding, and faster recovery . Biopsy therapy, one of the commonest source of cadavers is limited, and from the perspective of MIS, is a kind of needle insertion medical procedure, assisted the concept of protecting animals, animals used for surgical by a real-time medical image (ultrasound/CT), which navi- training are also subject to ethical restrictions. For inexperi- gates the needle percutaneously inserted through the human enced medical personnel, multiple surgical training is inevi- trunk to reach the lesion target. Since the needle insertion table, especially for diﬃcult operations, such as the lung, goes through diﬀerent tissue layers, such as the skin, fat, liver and other related operations. Therefore, it is important and muscle, to ﬁnally reach an internal organ [2, 3], it is hard to provide a good surgical training platform for the inexperi- to handle the puncture position angle and force during the enced medical personnel and interns. Surgical procedure operation; consequently, MIS has strict requirements for sur- simulations and robot-assisted surgery provide a good train- geons: they must undergo a long training period to gain sur- ing platform for MIS training. Virtual reality (VR) emerged gical experience, which means the curve of surgical learning as an essential technology of medical education applications on this minimally invasive surgery is extremely steep. Inexpe- in recent years; beneﬁting from these cutting-edge facilities, rienced surgeons and interns usually use animals, artiﬁcial surgical procedure simulations and robot-assisted surgery 2 Applied Bionics and Biomechanics iﬁcation showed that the insertion simulation can locate the have become increasingly immersive and reliable by adding visual and haptic feedback rendering [4–6]. Nevertheless, needle tip into the target accurately. Menciassi et al.  have the challenge of accurate and realistic haptic feedback cannot proposed a novel instrument and method to measure in vivo be overlooked, due to the high-frequency update require- biomechanical properties of tissue. In addition, Han et al. ment (1000 Hz). The big problem is the contradiction  developed an ultrasound indentation system to eﬃ- between accuracy and real time; to improve the accuracy of ciently measure soft tissue mechanical properties in vivo. In the system, it is necessary to make appropriate sacriﬁces in the second species, many researchers focus on modeling real time; many existing force calculation models based on insertion force; Simone and Okamura  took an interest either artiﬁcial or animal-based experimental data cannot on in vitro experiments on bovine livers using the Johns truly reﬂect the mechanical behavior response of soft tissues, Hopkins University Steady Hand Robot for percutaneous thus failing to meet the force feedback authenticity require- therapy, and the needle insertion forces were modeled based ments [7, 8]. In this paper, we proposed an immersive on the properties of soft tissue. The insertion forces were dynamic haptic sensation model during biopsy therapy with divided into three components: capsule stiﬀness, friction, human respiratory motion evaluated by intraoperative clini- and cutting. Okamura et al.  also presented a model cal data. The major contribution of this work is the higher ﬁt- for insertion force and experiment for acquiring data from ting degree force modeling based on real-time clinical data. ex vivo liver tissue; the insertion process has two phases: In addition, a dynamic biomechanics experimental frame- prepuncture and postpuncture, where the stiﬀness force work is also designed to provide higher speciﬁcation and an belonged to the former and friction force and cutting force immersion haptic rendering scenario. Last but not least, the are the latter. The experiment was performed to describe proposed VR-based biopsy training evaluation simulator the eﬀects of friction force and needle geometry during the complements the training immersion and minimizes the robotic needle insertion into tissue. Moreover, the experi- gap between virtual and reality. To the best of the authors’ ment setup was constructed for friction tests under the CT knowledge, it is innovative to integrate intraoperative surgi- scanning imaging. Some proposed a new puncture object to cal data into dynamic haptically enabled real-time VR medi- replace the animal soft tissue . Ng et al.  found an cal training simulation. appropriate method to model the interaction forces during The structure and content of this article are organized as needle insertion into porcine back tissue and simulated mul- follows: we brieﬂy make an introduction on biopsy surgery tilayer gelatin; the porcine sample was performed in vitro to and the challenges of VR-based medical training in the collect force data. The insertion force with simulated multi- layer gelatin compared with porcine sample shows that the above-mentioned part; after that, we review the previous related work on needle insertion biomechanics modeling. simulated multilayer gelatin can simulate the cutting force. Thirdly, the biomechanics data-based force model, with Thus, its insertion force resembles the actual porcine experi- clinical and VR training evaluation experiments, is designed ment and provides a suitable alternative medium for design in Methodology. Fourthly, the experimental ﬁtting and eval- and training needle insertion. And other people studied the eﬃciency factor of insertion forces. Jiang et al.  intro- uation results were demonstrated and discussed. Finally, we summarize the results and the potential contributions this duced the mechanism and inﬂuencing factors of the interac- paper makes. tion between puncture needle and soft tissue in detail; the result showed that the eﬀect of force has lots of factors, for example, needle geometry and soft tissue type. Gessert et al. 2. Related Work proposed a new method for needle tip force estimation by Currently, lots of researchers have studied the needle using OCT ﬁber imaging, and an ex vivo experiment with a insertion and proposed many mechanical models based on human prostate was tested . Apart from that, lots of the experimental data. The related work can be classiﬁed into researchers focus on realizing the haptic feedback during two species according to data acquisition: in vivo and in vitro. needle insertion by using some of the existing commercial force feedback equipment, such as the PHANTOM haptic In the ﬁrst species, Brouwer et al.  and Brown et al.  carried out the experiment from in vivo soft tissue to analyze device . In 2017, Xue et al. presented a teleoperated nee- dle insertion system with haptic feedback; they designed a the properties of soft tissue and describe the insertion force proﬁle. For force modeling, Maurin et al.  performed a novel force control framework to match the force signals, series of experiments; they compared the maximum force and the relationship between motor current and feedback force is modeled . In 2018, Yang et al. summarized the of diﬀerent organs during insertion in vivo and even gave a conclusion about the diﬀerence between manual and robotic. key components surrounding the force feedback and control Barbé et al.  found a method to estimate online the needle during robot-assisted needle insertion; this review involved forces with a low velocity of the needle tip and analyzed force modeling, identiﬁcation, and feedback control of the modeling and the interaction between the needle and tis- robot-assisted needle insertion . According to the reviewed literature above, it can be sum- sue during in vivo experiments. Moreover, researchers also developed a simulation system for insertion and measured marized that two major ﬂaws will appear when integrating the biomechanical properties of in vivo soft tissue. Kobayashi these models with the biopsy simulation: most of the experi- et al.  studied an integrated robot-assisted system with mental data recording was performed under the static condi- ultrasound-guided and simulation soft tissue deformation tion, without the respiratory motion; however, the breath action drives the human viscera up and down throughout using in vivo experimental data on the porcine liver; the ver- Applied Bionics and Biomechanics 3 Table 1: Parameters of tissue movement during respiration. Kidney (left) Heart Lung Liver Dorsal muscles Motion (mm) 7.1 (4.5-9.8) 5.2 (2.4-7.9) 8.6 (5.2-12.0) 4.5 (3.0-6.0) 2.9 (0.3-5.5) Breath cycle (s) 5 (4-6) Figure 1: The entire simulation system of the biomechanics experiment. 1: acquisition software; 2: multichannel data conversion card; 3: servomotor controller; 4: servomotor actuator; 5: linear guideway (vertical); 6: connector; 7: force sensor (ATI Nano17); 8: surgical trocar needle; 9: biopsy sample; 10: linear guideway (horizontal); 11: power supply. the biopsy procedures. Moreover, there is no doubt that The biomechanics test platform mainly consists of the experimental force data from in vivo is more reliable than surgical instrument, force sensor acquisition system, stepper in vitro, but are these animal experimental data-based force motor drive system, container for soft tissue, and computer, models suited for the haptic rendering in the VR-based train- which are detailed and demonstrated in Figure 1. The bio- ing simulator of human biopsy surgery? With respect to these mechanics data is collected from 5 diﬀerent part tissues of points, we conducted the solutions with a dynamic biome- porcine (in Table 1), which are bought from the butcher chanical experiment platform and evaluated with clinical market within 24 hours. The experiment temperature is data in the following parts. 19 ± 0 5 C, humidity is 25%, and the biopsy needle parame- ter is 18G trocar with a 1.06 mm inner diameter and 15 cm long. We select ATI Nano17 as the force sensor; it is installed 3. Methodology on the handle of the trocar to collect the information of the insertion force in real time; the precision is 0.01 N. The step- 3.1. Biomechanics Data-Based Force Model per motor is mainly composed of a slider, subdivision, and controller, the maximum motion range is 200 mm, motion 3.1.1. Biomechanics Test Platform. Respiratory motion is an unavoidable inﬂuencing factor of the needle insertion preci- accuracy is 0.16 μm, and the velocity can be programmed to sion during the biopsy surgery; nevertheless, most of the for control, with the controllable range of 1 mm/s-30 mm/s. There are two stepper motor drive systems: one is used to researchers focus the study on the static puncture in vitro. Force is recorded in the ideal without-breath condition, control the vertical velocity of the biopsy needle insertion regardless of the movement of the organ by the respiration and the other one is used to simulate the human respiratory during the operation. To resolve this issue, this study motion to achieve dynamic puncture. The literature sug- designed a novel biomechanical platform to record the gested that the velocity of biopsy needle insertion is usually in the range of 0.1 mm/s~10 mm/s ; therefore, we set dynamic biopsy force with respiratory motion. According to Weiss et al.’s research , the primary motion of human the needle insertion with the constant velocity of 3 mm/s in this study. The size of the container is 12 mm × 7 mm × organs’ 3D centroid is the head-foot direction movement; based on the 4D-CT data, the tissue has simple motion 7mm, and the sampling frequency of the force sensor is ranges [25–27] and the parameters we chose in this work 1000 Hz. Apart from that, the experimental platform must be placed on the horizontal table to avoid the chattering are demonstrated in Table 1. 4 Applied Bionics and Biomechanics phenomenon during needle insertion; the needle tip and the yx = α + 〠 α cos nωx + β sin nωx , soft tissue surface are in a vertical state and have a distance, 0 n n n=1 which ensures the trocar can transform from acceleration to the uniform motion. The whole experimental setup is T α = yx dx, shown in Figure 1. In order to reduce the experimental error, each kind of tissue experiment was performed on three dif- ferent regions and then the average value is calculated; only α = yx cos nωx dx, n =1,2,⋯∞, the average value is determined as the analysis data for the force modeling. 3.1.2. Insertion Force Modeling. The mechanical properties of β = yx sin nωx dx, n =1,2,⋯∞, soft tissues are complex and usually exhibit characteristics such as nonlinearity and anisotropy . The study of their −inωx mechanical properties is mainly to analyze the stress-strain m = yx e dx, n = 0,±1,±2,⋯∞ 7 relationship, but there is no clear mechanical model to deﬁne it strictly because the internal structure of the soft tissue is Based on the above formula, the superposition coeﬃ- complex; most of them are based on the properties of soft tis- cient of a function expression with a known periodic func- sue and experimental data to analyze the mechanics model- tion can be obtained by integration. In our simulation, ing. Simone and Okamura  used a nonlinear spring since the force model is unclear, we employed the function model as a model for the stiﬀness and a modiﬁed Karnopp approach method to ﬁt the experimental data: model as a model for the friction; the cutting was obtained by subtracting friction from the whole forces; it can be con- sidered as a constant for a given material. Okamura et al. lim yx − α + 〠 α cos nωx + β sin nωx dx =0 0 n n N→∞  found that it was best to ﬁt the data by a second-order n=1 polynomial for prepuncture. And Maurin et al.  analyzed the experimental data using an exponential function of the depth in each puncture phase. Some scholars assume that According to this, we introduced ﬁnite triangular the soft tissue has a linear characteristic and propose a linear periodic functions approaching the puncture force function mechanical model, but this does not meet the objective y x to describe the ﬁnal model as requirements and the actual needs. As far as we know, the needle puncture procedures are continuity, uncertainty, and independence, which may be yx = α + 〠 α cos nωx + β sin nωx 9 0 n n n=1 hard to describe with the existing model. Therefore, it is nec- essary to ﬁnd a mechanical model that can be applied to both α and β are the Fourier coeﬃcients, and N means the the puncture process analysis and the general application. As n term number of the Fourier series. In Figure 2, the force a powerful function, Fourier is usually used in various ﬁelds; model is analyzed through frequency spectrogram to deter- consequently, Fourier is applied to insertion force modeling mine the Fourier series N, which is set from 1 to 8. in this paper. In order to incorporate the Fourier model into the mechanical modeling of percutaneous biopsy, we intro- 3.2. Clinical Evaluation Experiment Design duced the periodic extension idea to convert the nonperiodic puncture force into a periodic function while extending the 3.2.1. Clinical Data Collection Platform Design. We chose the function deﬁnition domain to the entire real ﬁeld. Because renal biopsy surgery as the clinical data validation procedure the real-time puncture force is centered on the low- which is an invasive procedure to obtain a small piece of frequency part, according to Fourier deﬁnition, each periodic tumor sample from the renal tissue for pathologic examina- function can be decomposed of numbers of trigonometric tion. Clinical data collection trials are set up in the Division periodic functions, which is the Fourier series. The equation of Urology in Yunnan First People’s Hospital, Kunming, is as follows: Yunnan, China. The needle insertion usually can be divided into prepuncture and postpuncture; before insertion, the patient lies in the supine position, then the operator injects y x = α cos nωx + β sin nωx 1 local anesthesia into the skin, through the subcutaneous tis- sue and down to around the kidney. After the preliminary work is prepared, a biopsy is performed by the operator with ix Because of the Euler formula: e = cos x + i sin x,we the aid of real-time medical imaging for guidance. It provides transformed the trigonometric function into an exponen- critical information for clinicians to determine the condi- tial form: tion, treatment prescription, and postoperative estimation. In these procedures, what it is desired is more accuracy of ∞ insertion, reduced operation time, and less damage to sur- inωx 2 rounding tissue during the insertion. Because of one of the yx = 〠 m e , n=−∞ coauthors who is also a professional surgeon in the hospital, Applied Bionics and Biomechanics 5 1.5 0.5 0123456789 10 02468 10 12 Time (s) Frequency (Hz) Exp 1 Exp 3 Exp 2 Average Figure 2: The insertion force of the kidney and corresponding frequency spectrogram. 2 3 Figure 3: Clinical trial setup for the multichannel intraoperative data collection. Component details are as follows: 1: biopsy surgical instrument; 2: force sensor (ATI Nano17) +3D print connector; 3: line puncture device integration; 4: data conversion; 5: analysis software; 6: patient. all clinical and data collection procedures referring to the to conduct the percutaneous surgery while real-time force patients and the surgery are anonymous and nonidentiﬁable. data is recorded. The procedure involves a force acquisition Clinical procedures are planned and conducted by profes- component (ATI Nano17) and a guide image recording sional urologists in the hospital and in accordance with asso- component (ultrasound), as well as an audio recording com- ciated safety and ethical approvals. The clinical data adopted ponent. The force acquisition component is implemented in this paper was provided to the authors through a public based on a standard trocar needle (18G COOK trocar nee- link to a database on haptic force feedback (http://civ.ynnu dle), which is a ﬂexible surgical instrument, consists of a nee- .edu.cn/ChineseShow.aspx?ID=1). Figure 3 shows the clini- dle core and needle sheath which is demonstrated in Figure 4, cal system setup for clinical data collection and the detailed and is speciﬁcally for such procedures. A force sensor, a data system components in the surgical room implementation. collector, and a power supply were also integrated with the trocar needle, as shown in Figure 4. The force sensor is an 3.2.2. Experiment Setup. The actual percutaneous therapy ATI Nano17. It connects to the rear end of the trocar needle through a customized 3D print connector. involves a urologist holding the percutaneous instrument Force (N) |A(f)| 6 Applied Bionics and Biomechanics Figure 4: Percutaneous intraoperative force recording instrument includes a 6 DoF force sensor, a biopsy needle, and a customized 3D printed connector. Table 2: Renal biopsy components in the operation room and the CT dataset provided by the Yunnan First People’s Hospital, virtual simulator. Kunming, China. The dataset has a dimension of 512 × 512 × 3172 and 0 51 × 0 51 × 0 50 mm accuracy, as demon- Operation room Virtual simulation Number strated in the second line of Figure 6. The CT dataset (Figure 4) (Figure 5) (DICOM format) was ﬁrst imported for ROI extraction using 1 Guide image recording Original guide image segmentation functions such as threshold and region grow- 2 Audio recording Hi-ﬁ headphone ing. Afterwards, 3 urologists were invited from Yunnan First 3 Ultrasound head Ultrasound head People’s Hospital to validate the autosegmentation result 4 ATI Nano17 PHANTOM Omni and perform manual correction where necessary, as demon- 5 18G trocar needle 18G trocar needle strated in the second line of Figure 6. Finally, we employed the marching cube algorithm to reconstruct the dataset into 6 Patient CT reconstruction a 3D mesh model, with redundant mesh cleaning and Lapla- 7 Force output Force rendering cian smoothing as intermedium steps. Finally, we recon- 8 Surgical environment HTC VIVE structed the whole surgical environments in MAYA and added the background audio inside into the scene render- As the CT images used for visual reconstruction could be ing shown in the third line of Figure 6. Two PHANTOM obtained before the operation, only the guided image record- Omni reproduce the dual-hand force rendering, HTC ing is required in real time. Ultrasound images were recorded VIVE provides the 3D visual rendering with head tracking, with a screen capture software as a guiding image. Each and a Beat Solo 2 wireless headphone provides the real- frame was recorded for the guide image in the proposed time surgical audio. training environments. The recorded images are cropped to the preoperation part as the training interface needs to be 3.3.2. System Evaluation Design. 18 novice and 9 expert overlaid with only a shadow’s projection of the needle. On urologists are invited to assess the virtual reality training the audio recording part, we adopted a Sony PCM-D100 platform. Before the training session starts, a didactic intro- high-resolution recorder to capture the working audio of duction of system operation is provided to both the expert machines inside a theatre. We then replay these as the back- and novice groups. Furthermore, a customized-built Global ground audio in our training system. Table 2 shows the Rating Scale (GRS) questionnaire is introduced for the sub- biopsy surgical components in Figure 4 and the correspond- jective evaluation [29–31]. The time interval of an objective ing components in a virtual simulator in Figure 5. test is 30 min between the pretest and posttest groups of the novices. The time interval between the training sessions is 24 hours. After ﬁnishing the experiments, all 27 trainees’ per- 3.3. VR-Based Training Evaluation Simulator Design formance was recorded by the subjective questionnaires to 3.3.1. Visual and Haptic Rendering Pipeline. In the proposed evaluate our biopsy therapy training system with respiratory training framework, the urologist guides the real-time C- motion. The evaluation system procedures are demonstrated ARM or ultrasound image navigation to locate the minimal in Figure 5. invasive position for needle insertion. To provide an immer- Comparisons between groups (novice–expert) were ana- sive haptic rendering of such dual-hand operation, we imple- lyzed using the Mann–Whitney U test (continuous variables) mented both ultrasound force and needle insertion force or the chi-square test (categorical variables) of the stored data within one haptic rendering loop. The ﬁrst line in Figure 6 in our system [6, 32]. Pretraining and posttraining data in the demonstrates the proposed two diﬀerent haptic rendering 2 groups were compared using paired t-tests. Spearman rank schematic, which could potentially be used in many other order correlations were calculated to determine correlations image-guided medical simulations. Visual rendering is between objective simulator-derived scores and the GRS responsible for visual clue setup according to an anonymous results. P value < 0.05 was considered statistically signiﬁcant. Applied Bionics and Biomechanics 7 Figure 5: The immersive virtual training platform with an i7 6700 (3.4 GHz) CPU workstation, 8 GB memory, and 1070 NVIDIA Graphics GPU. Two PHANTOM Omni reproduce the dual-hand force rendering, HTC VIVE provides the 3D visual rendering with head tracking, and a Beat Solo 2 wireless headphone provides the real-time surgical audio. Component details corresponding with the operation room are demonstrated in Table 2. Analyses were performed using the software package IBM People’s Hospital in Kunming; all the procedures related SPSS Statistics, version 20.0 [33, 34]. The evaluation system to the patients and the operation are designed and con- procedure design underwent our immersive virtual percuta- structed by the professional surgeons according to the safety neous therapy simulator as shown in Figure 7. and ethical rules. Based on the result in Figure 9 and Table 4, the proposed dynamic model in this study demonstrated a high ﬁtting performance with the interoperated data in a 4. Results real operation. 4.1. Results of Biomechanics Force Modeling. After the exper- iments, the data were analyzed by MATLAB 2016 software. 4.3. Evaluation Results of the VR-Based Training Simulator. The nonlinear least squares were utilized to ﬁt the insertion Each expert participant could perform the puncture attempts force by using the Fourier model proposed in Section 3.1.2. on the model and successfully hit the provided tumor in To avoid coincidence, three diﬀerent regions of the same the calyx. The survey showed that 7 participants (77.8%) would use our framework to train novices and 8 urologists individual are performed for comparing the tendency of insertion force; ﬁnally, the average value was calculated and (88.9%) considered the kidney model as an accurate ana- then the insertion force was modeled except the retraction tomic representation with only 1 participant (11.1%) who phase. The relationship curve between the insertion force did not agree with this statement. 6 participants (66.7%) and the time is shown in Figure 8; as can be seen from the ﬁg- considered the depicted the graphic simulation is as real as the real surgical scene. The other 3 were neutral about this ure, the tendency of the insertion force has broad resem- blance but the position of the peaks and valleys of the statement. 5 (55.5%) participants thought the system had a insertion force has a little deviation due to the anisotropy high performance of haptic feedback; 2 (22.2%) of them were and inhomogeneity of the internal tissue of the soft tissue. neutral. 8 attendants (88.9%) considered the X-ray-simulated Therefore, for the same object, there is a tiny diﬀerence in image as an accurate representation of a real ﬂuoroscopic image, and all of them (100%) felt that avoiding radiation the data curve for diﬀerent regions; mechanical modeling can be used in the same model . was important for training; the detailed score of GRS and A detailed piecewise ﬁtting degree of ﬁve kinds of tissues the objective evaluation on the VR-based training simulator are shown in Table 5. is demonstrated in Table 3; parameters of tissue movement during the respiration are set based on Table 1. For the system subjective questionnaire evaluation result, novices’ appraisement demonstrated signiﬁcantly 4.2. Evaluation Results of the Haptic Model by Clinical Data. higher than experts’ in total performance. Details of the sub- To verify the modeling method, the clinical trial setup is jective evaluation of the VR-based training simulator are shown in Table 6. conducted in the Urology Department of the Yunnan First 8 Applied Bionics and Biomechanics Mechanical Force Haptic recording data rendering 3D Medical Visual images reconstruction rendering Scene Surgical 3D rendering environment reconstruction (a) (b) (c) Figure 6: The implementation of the medical training platform with direct intraoperative data, which includes the multichannel data recording in the operation room (a), data processing and reproduction (b), and ﬁnal virtual medical training environment reconstruction (c). Haptic rendering Visual rendering Patient CT images Biomechanics data 2D segmentation Force modeling 4D reconstruction Clinical data validation Visual model Haptics model 9 experts 18 novices GRS score Objective GRS score Objective Questionnaire Questionnaire Expert group Novice group Figure 7: The evaluation system procedure design underwent our immersive virtual percutaneous therapy simulator. 5. Discussion require the needle insertion into a speciﬁc part of the diseased area as precise as possible . For cytopathological exami- As medical surgery, percutaneous operation is often uti- nation, with the navigation of intraoperative images, resi- dents usually rely on their clinical experience to manipulate lized for cancer treatment, and cytopathological examina- tion, as the cancer treatment method, such as percutaneous the operation to extract the lesion sample. Both are widely ethanol injection therapy (PEIT) and radiofrequency abla- used in the invasive procedure. Like Gordon et al.  based tion (RFA) is performed for liver cancer; both of them on the MATLAB program, the data was segmented into Presimulation Intrasimulation Applied Bionics and Biomechanics 9 1.5 1.5 1 1 0.5 0.5 −0.5 0 2 4 6 8 10121416 0 3 6 9 12 15 18 21 24 27 Time (s) Depth (mm) 3.5 2.5 1.5 0.5 −1 0 0123456789 10 −0.5 0 3 6 9 12 15 18 21 24 Time (s) 1.4 Depth (mm) 1.2 1.2 0.8 0.6 0.8 0.4 0.6 0.2 0 0.4 −0.2 0.2 02468 10 12 Time (s) 0 5 10 15 20 25 30 35 40 4.5 Depth (mm) 4.5 3.5 3 3.5 2.5 2.5 1.5 1.5 1 1 0.5 0.5 0 5 10 15 20 25 −0.5 Depth (mm) 1.5 Time (s) 1.2 1.5 0.9 0.6 0.3 0.5 0369 12 15 18 21 24 27 30 0123456789 10 Depth (mm) Time (s) Actual Exp 1 Exp 3 Exp 2 Average Model (a) (b) (c) Figure 8: The biomechanics force data ﬁtting results: tissue sample (a), initial data and average (b), and mean-value force modeling with our proposed force model (c). piecewise regions and modeled. In this work of the insertion ical model proposed in this paper has a good eﬀect on the force model, the Fourier series N was adjusted according to insertion force. In addition, a series of needle insertion exper- the diﬀerent regional characteristics, and then the model iments were performed to verify the applicability of the Fou- parameter was estimated by the algorithm. We have some rier model. Next, needle insertion was performed on four information from the ﬁtting curve, the Fourier model can diﬀerent tissues: the liver, lung, heart, and pork. As with kid- describe the experimental data well, and it has a good coinci- ney insertion, each group experiment tested three diﬀerent dence with the experimental data, especially the ﬂuctuation regions and calculated the average value, and each type of phenomenon in the process of needle insertion. The mechan- experimental data was analyzed and modeled using the Force (N) Force (N) Force (N) Force (N) Force (N) Force (N) Force (N) Force (N) Force (N) Force (N) 10 Applied Bionics and Biomechanics Table 3: Fitting degree between the biomechanics data and our Table 6: Results of subjective evaluation on the VR-based training force model. simulator. Phase Fitting degree Subjective questionnaires Expert (9) Novice (18) Type 1 2 3 4 5 Average Dynamic force model eﬀects (range 1-5) Kidney (left) 0.9975 0.9446 0.9939 0.9936 0.9980 0.9855 Realism of respiratory 3.7 3.9 Liver 0.9996 0.9988 0.9998 0.9994 0.9999 0.9995 Realism of haptic feedback 3.9 4.2 Lung 0.9997 1 1 0.9979 0.9993 0.9994 Realism of needle insertion 4.1 4.1 Heart 1 0.9999 1 0.9310 0.9999 0.9862 Haptic fatigue 2.3 1.4 Dorsal muscles 0.9997 0.9979 0.9999 0.9999 0.9996 0.9994 System evaluation (range 1-5) VR graphic performance 3.9 4.2 Surgical tool manipulation 3.4 3.7 Visual fatigue 1.1 1.2 Comfortable HMC training 3.7 3.5 Hand-eye coordination training 4.1 4.5 Overall appraisal 3.6 3.4 Exceptions? (range 1-5) Surgical navigation 3.1 3.5 −1 0369 12 Surgical planning 4.2 4.7 Depth (mm) Novice training 4.7 4.6 Actual Resident practice 3.9 4.1 Model Surgical rehearsal 3.5 3.8 Figure 9: The ﬁtting result of intraoperative data using our force model. high degree of ﬁtting for insertion force, it is not accidental, and it is also applicable to each tissue type. And it is further Table 4: Fitting degree between clinical data and our force model. illustrated that the Fourier model has a generalization. More- Phase 1 2 3 4 5 Average over, there is obvious information according to the above experimental data of four tissue types: Fitting degree 0.9768 0.9953 0.9991 0.9980 0.9990 0.9936 (i) For diﬀerent tissue types, the stiﬀness force and the Table 5: Results of GRS and objective evaluation on the VR-based force distribution are diﬀerent training simulator. (ii) For the same tissue type, the insertion force curves Experts (9) Novice (18) P value vary from region to region, but the tendency of the GRS (range 1-5) force curve is similar; only the force peaks and slopes have a little diversity 2 8± 013 2±0 2 Identify anatomy <0.001 2 2± 023 1±0 2 Plan needle puncture <0.001 (iii) The complexity of the organizational structure dis- tribution is known from the shape of the mechanical 2 1± 013 1±0 1 Instrument use <0.001 curve by observing the ﬂuctuation of the force curve 2 3± 013 4±0 2 Ability to perform tasks <0.001 during needle insertion. The distribution of the 2 5± 013 6±0 2 Overall performance <0.001 internal structure was simple when the ﬂuctuation Objective assessment was calm, but vice versa 6 8± 048 9±0 5 Operation time (min) <0.001 In this paper, the experimental data were modeled using 6 2± 038 6±0 3 Fluoroscopy time (min) <0.001 the above classical methods and the Fourier model proposed 2 4± 023 8±0 2 No. needle punctures 0.002 by us. As shown in Figure 10, it is obviously seen that the Fourier model has a better coincidence with the experimental 0 4± 0 08 1 2±0 1 Infundibular injury <0.001 data than the other models, especially when the insertion 0 5± 011 8±0 1 PCS perforations 0.024 force ﬂuctuates; the polynomial model and the nonlinear 1 5± 022 7±0 2 Rib injury 0.02 model have diﬃculty in describing the tendency of insertion 2 1± 033 7±0 3 Blood vessel injury 0.07 force. Table 7 shows the accuracy of each model to ﬁt the insertion phase; it can be also seen from Table 7 that the Fou- rier model shows a high performance on the experimental Fourier model presented in this paper. The following are the data; the average of the ﬁtting precision is 0.9855. Conse- result of data acquisition and modeling for each tissue type. quently, in this paper, we employed the Fourier model to analyze the whole needle insertion phase. From the above modeling results, the Fourier model has a Force (N) Applied Bionics and Biomechanics 11 1.8 Table 8: Fitting results between existing ﬁtting algorithms with 1.6 clinical data (renal biopsy). 1.4 1.2 Phase The ﬁtting degree Model 1 2345 Average 0.8 0.6 Polynomial 0.6248 0.8738 0.9390 0.9558 0.9980 0.8783 0.4 Nonlinear 0.6269 0.8420 0.9365 0.4726 0.7254 0.7207 0.2 Exponential 0.4076 0.8548 0.9375 0.3017 0.8331 0.6670 −0.2 0 5 10 15 20 25 Our method 0.9768 0.9953 0.9991 0.9980 0.9990 0.9936 Depth (mm) Experimental data Nonlinear model clinical data, what one should pay more attention to is the Fourier model Exponent model Polynomial model discontinuity and unsmoothness between nodes in the piecewise modeling; it is a deﬁciency of the Fourier model Figure 10: Comparison with the existing ﬁtting algorithms of and must be modiﬁed. In the past years, the mechanical experimental data (left kidney). properties of biological soft tissues were studied through mechanical experiments. In recent years, research trends Table 7: Fitting results between existing ﬁtting algorithms with use theoretical models to analyze and ﬁt experimental experimental data (left kidney). results. That is to say, a mathematical expression is used to describe the mechanical behavior of soft tissue by combining Phase The ﬁtting degree the theoretical model with experimental data. For simulating Model 1 2 3 4 5 Average the puncture biopsy accurately, the next work is going to Polynomial 0.9934 0.6981 0.9573 0.8382 0.9463 0.8867 ﬁnd a mechanical model considering the viscoelasticity and Nonlinear 0.9863 0.5190 0.9022 0.7791 0.7555 0.7880 hyperelasticity of soft tissue. Moreover, collecting and Exponential 0.9770 0.5511 0.8644 0.6810 0.9028 0.7953 enriching the database of soft tissue systematically will be Our method 0.9975 0.9446 0.9939 0.9936 0.9980 0.9855 an essential step in the future for the establishment of big data medical analysis. To sum up, the principal contribution of this paper is to implement and verify the modeling method, by obtaining the clinical data of the needle puncture and comparing with the experimental data of a dynamic framework, which illus- trated that the dynamic insertion model can eﬀectively simu- late the real operation. For the system evaluation result, the face and content appraisal median value was 4 (1-5) by 9 experts, which is −1 the same score as the existing percutaneous biopsy surgical 02468 10 12 training platform. Experts’ performance is signiﬁcantly Depth (mm) higher than novices’ in total performance. Posttest values of Experimental data Nonlinear model novice group after a few training sessions have shown signif- Fourier model Exponent model icant improvement with respect to pretest GRS and objective 2nd polynomial scores. In all, Table 6 demonstrates that the proposed simula- model tion platform’s evaluation showed higher performance than Figure 11: Comparison with the former ﬁtting algorithms of the existing benchmark training platform. Simulation com- clinical data (renal biopsy). plexity and graphics on visualization showed a higher score with the novice students; this can be explained by the fact From Figure 11 and Table 8, we can see that the model that the real ultrasound images are integrated into the oper- this paper proposed can eﬀectively simulate the interoper- ation face which provides an accurate visual indication for ated data in a real operation. In addition, the experimental trainees. On the other hand, the training tools and assess- ment tool displayed a better performance with the novice; data of the kidney was utilized to compare with the intraop- erative data; we ﬁnd from the similarity from Figure 8 that this result may have arisen by the fact that the needle simula- there is a range of ﬂuctuations in the force during needle tor in the current version utilized the real surgical needle insertion whether it is a manual insertion or a robot hand integrated with a low-ﬁdelity stylus of the haptic device of insertion; this phenomenon could have been contributed to the simulator, which may be confusing for the experts; more- over, the missing foot pedal ﬂuoroscopy controls may have by the soft tissue’s material properties. But there is a diﬀer- ence between the intraoperative data of the manual control also led to the score deduction. Most of the experts still had and the experimental data of the robot, in which the insertion doubts about considering it as an assessment tool and would force of the manual control is greater than the insertion force prefer further validation studies to reach a conclusion. of the robot. Although the model of the insertion force we Besides the subjective validation on the face and content, more objective evaluations also require construct validity. proposed shows great work in ﬁtting experimental data and Force (N) Force (N) 12 Applied Bionics and Biomechanics Figure 12: Real-time biopsy therapy VR training simulator with respiratory motion (left kidney). To our knowledge, this study is the ﬁrst time to validate a tively apply to several tissue types; it is further veriﬁed that VR-based biopsy simulator by utilizing dynamic original the mechanical model proposed in this paper has a certain clinical haptics, which provided a high-ﬁdelity simulation generalization. The future work will include the following: of the biopsy construct validity in Table 5. Experts had sig- the experimental equipment needs to be optimized, such as niﬁcantly shorter resection times than novices. In addition, the simulation of human respiratory system, and experi- our simulator demonstrated a higher score in the “identify ments need to be performed to enrich experimental data anatomy” and “overall performance.” The rib injury number and types. in our simulator is much lower than that in the existing sim- ulator [36, 37]; the reason should be the facilitation of the Data Availability guiding hand which can detect the rib before puncture. The experimental data used to support the ﬁndings of Blood vessel injury is much higher and without statistical this study are available from the corresponding author signiﬁcance because of the reconstruction model from the upon request. spiral CT which includes many blood capillaries; however, many of these vessel injuries can be ignored during the biopsy. The real-time respiratory motion eﬀect during the Conflicts of Interest VR biopsy simulator is demonstrated in Figure 12. The authors declare that there is no conﬂict of interest regarding the publication of this paper. 6. Conclusion Acknowledgments In this paper, a novel force modeling methodology for the biopsy therapy with respiratory motion using direct clinical This research is funded by the National Natural Science data is proposed; the dynamic biopsy biomechanics experi- Foundation of China (61650401) and the Natural Science ment architecture is designed and constructed to facilitate Foundation of Yunnan Province, China (ZD2014004) of the accurate modeling of the insertion force. In addition, Yunnan Key Laboratory of Opto-electronic Information clinical evaluation is conducted and performed, and VR- Technology, Kunming, China. We thank Professor Minghui, based training evaluation based on the proposed haptic ren- Xiao, Dr Min Zou, and Dr. Jie Shen of Yunnan First People’s dering solution is also performed, which demonstrated the Hospital for the helpful surgical suggestions. validity of the mechanical model and the feasibility of the evaluation. It is the ﬁrst time to propose a dynamic force References modeling and introduce the idea of continuations modeling of real-time biopsy force rendering. The principal contribu-  E. P. Westebring-van der Putten, R. H. M. Goossens, J. J. tion of this paper is to establish a complete system for the Jakimowicz, and J. Dankelman, “Haptics in minimally inva- study of percutaneous biopsy training, including experimen- sive surgery – a review,” Minimally Invasive Therapy & Allied tal construction, theoretical analysis, and surgical laboratory Technologies, vol. 17, no. 1, pp. 3–16, 2008. validation and evaluation, which is ﬁnally based on the VR  A. Carra and J. C. Avila-Vilchis, “Needle insertion modeling training simulation assessment, which is a comprehensive through several tissue layers,” in 2010 2nd International Asia research and veriﬁcation process. The Fourier model was Conference on Informatics in Control, Automation and Robot- used to model the experimental data of diﬀerent tissue types, ics (CAR 2010), vol. 1, pp. 237–240, Wuhan, China, March which illustrated that the dynamic force modeling can eﬀec- 2010. Applied Bionics and Biomechanics 13  K. W. Ng, J. Q. Goh, S. L. Foo, P. H. Ting, and T. K. Lee, “Nee-  N. Abolhassani, R. Patel, and M. Moallem, “Control of soft tis- sue deformation during robotic needle insertion,” Minimally dle insertion forces studies for optimal surgical modeling,” Invasive Therapy & Allied Technologies, vol. 15, no. 3, International Journal of Bioscience, Biochemistry and Bioinfor- pp. 165–176, 2006. matics, vol. 3, no. 3, pp. 187–191, 2013.  D. Fortmeier, A. Mastmeyer, J. Schroder, and H. Handels,  S. Jiang, P. Li, Y. Yu, J. Liu, and Z. Yang, “Experimental “A virtual reality system for PTCD simulation using direct study of needle-tissue interaction forces: eﬀect of needle visuo-haptic rendering of partially segmented image data,” geometries, insertion methods and tissue characteristics,” IEEE Journal of Biomedical and Health Informatics, vol. 20, Journal of Biomechanics, vol. 47, no. 13, pp. 3344–3353, no. 1, pp. 355–366, 2016. 2014.  S. Y. Selmi, G. Fiard, E. Promayon, L. Vadcard, and J. Troccaz,  N. Gessert, T. Priegnitz, T. Saathoﬀ et al., “Needle tip force “A virtual reality simulator combining a learning environment estimation using an OCT Fiber and a fused convGRU-CNN and clinical case database for image-guided prostate biopsy,” architecture,” in Medical Image Computing and Computer in Proceedings of the 26th IEEE International Symposium on Assisted Intervention – MICCAI, vol. 11073, pp. 222–229, Computer-Based Medical Systems, pp. 179–184, Porto, Portu- Springer, Cham, 2018. gal, June, 2013.  R. Kokes, K. Lister, R. Gullapalli et al., “Towards a teleoperated  S. Ullrich and T. Kuhlen, “Haptic palpation for medical simu- needle driver robot with haptic feedback for RFA of breast lation in virtual environments,” IEEE Transactions on Visual- tumors under continuous MRI,” Medical Image Analysis, ization and Computer Graphics, vol. 18, no. 4, pp. 617–625, vol. 13, no. 3, pp. 445–455, 2009.  L. Xue, Y. Xie, K. Zhang, Y. Chen, and G. Geng, “A haptic force  S. Chan, P. Li, G. Locketz, K. Salisbury, and N. H. Blevins, feedback system for teleoperated needle insertion,” in 2017 “High-ﬁdelity haptic and visual rendering for patient-speciﬁc IEEE International Conference on Robotics and Biomimetics simulation of temporal bone surgery,” Computer Assisted Sur- (ROBIO), vol. 2018, Macau, China, December, 2017. gery, vol. 21, no. 1, pp. 85–101, 2016.  C. Yang, Y. Xie, S. Liu, and D. Sun, “Force modeling, iden-  A. Mastmeyer, D. Fortmeier, and H. Handels, “Evaluation of tiﬁcation, and feedback control of robot-assisted needle direct haptic 4D volume rendering of partially segmented data insertion: a survey of the literature,” Sensors, vol. 18, for liver puncture simulation,” Scientiﬁc Reports, vol. 7, no. 1, no. 2, p. 561, 2018. p. 671, 2017.  E. Weiss, K. Wijesooriya, S. V. Dill, and P. J. Keall, “Tumor and  I. Brouwer, J. Ustin, L. Bentley, A. Sherman, N. Dhruv, and normal tissue motion in the thorax during respiration: analysis F. Tendick, “Measuring in vivo animal soft tissue properties of volumetric and positional variations using 4D CT,” Interna- for haptic modeling in surgical simulation,” Studies in Health tional Journal of Radiation Oncology, Biology, Physics, vol. 67, Technology and Informatics, vol. 81, pp. 69–74, 2001. no. 1, pp. 296–307, 2007.  J. D. Brown, J. Rosen, Y. S. Kim, L. Chang, M. N. Sinanan, and  S. Siva, D. Pham, S. Gill et al., “An analysis of respiratory B. Hannaford, “In-vivo and in-situ compressive properties of induced kidney motion on four-dimensional computed porcine abdominal soft tissues,” Studies in Health Technology tomography and its implications for stereotactic kidney radio- and Informatics, vol. 94, pp. 26–32, 2003. therapy,” Radiation Oncology, vol. 8, no. 1, 2013.  B. Maurin, L. Barbe, B. Bayle et al., “In vivo study of forces dur-  Y. Hu, Y. K. Zhou, Y. X. Chen, S. M. Shi, and Z. C. Zeng, ing needle insertions,” in Perspective in Image-Guided Surgery, “4D-CT scans reveal reduced magnitude of respiratory liver Melbourne, Australia, July 2004. motion achieved by diﬀerent abdominal compression plate positions in patients with intrahepatic tumors undergoing  L. Barbé, B. Bayle, M. de Mathelin, and A. Gangi, “Needle helical tomotherapy,” Medical Physics, vol. 43, no. 7, insertions modelling: identiﬁability and limitations,” IFAC pp. 4335–4341, 2016. Proceedings Volumes, vol. 39, no. 18, pp. 129–134, 2006.  I. J. R. Oncology, S. Senan, C. E. Kleynen, B. J. Slotman, and  Y. Kobayashi, A. Onishi, H. Watanabe et al., “Development of F. J. Lagerwaard, “Renal Mobility during Quiet Uncoached an integrated needle insertion system with image guidance and Respiration: An Analysis of 4DCT Scans,” Materials/Methods, deformation simulation,” Computerized Medical Imaging and vol. 63, no. 2, pp. 210-211, 2005. Graphics, vol. 34, no. 1, pp. 9–18, 2010.  Y. C. Fung and R. Skalak, “Biomechanics: mechanical proper-  A. Menciassi, A. Eisinberg, M. C. Carrozza, and P. Dario, ties of living tissues,” Journal of Biomechanical Engineering, “Force sensing microinstrument for measuring tissue proper- vol. 103, no. 4, p. 231, 1981. ties and pulse in microsurgery,” IEEE/ASME Transactions on Mechatronics, vol. 8, no. 1, pp. 10–17, 2003.  P. Vávra, J. Roman, P. Zonča et al., “Recent development of augmented reality in surgery: a review,” Journal of Healthcare  L. Han, J. A. Noble, and M. Burcher, “A novel ultrasound Engineering, vol. 2017, 9 pages, 2017. indentation system for measuring biomechanical properties of in vivo soft tissue,” Ultrasound in Medicine & Biology,  A. Aydin, A. M. A. Shaﬁ, M. Shamim Khan, P. Dasgupta, and vol. 29, no. 6, pp. 813–823, 2003. K. Ahmed, “Current status of simulation and training models in urological surgery: a systematic review,” Journal of Urology,  C. Simone and A. M. Okamura, “Modeling of needle insertion vol. 196, no. 2, pp. 312–320, 2016. forces for robot-assisted percutaneous therapy,” in Proceedings 2002 IEEE International Conference on Robotics and Automa-  S. Mishra, A. Kurien, R. Patel et al., “Validation of virtual real- tion (Cat. No.02CH37292), vol. 2, pp. 2085–2091, Washington, ity simulation for percutaneous renal access training,” Journal DC, USA, May, 2002. of Endourology, vol. 24, no. 4, pp. 635–640, 2010.  A. M. Okamura, C. Simone, and M. D. O'Leary, “Force model-  K. Qian, J. Bai, X. Yang, J. Pan, and J. Zhang, “Essential ing for needle insertion into soft tissue,” IEEE Transactions on techniques for laparoscopic surgery simulation,” Computer Biomedical Engineering, vol. 51, no. 10, pp. 1707–1716, 2004. Animation and Virtual Worlds, vol. 28, no. 2, 2017. 14 Applied Bionics and Biomechanics  P. Korzeniowski, A. Barrow, M. H. Sodergren, N. Hald, and F. Bello, “NOViSE: a virtual natural oriﬁce transluminal endo- scopic surgery simulator,” International Journal of Computer Assisted Radiology and Surgery, vol. 11, no. 12, pp. 2303– 2315, 2016.  X. Chen, Y. Lin, C. Wang, G. Shen, and X. Wang, “A virtual training system using a force feedback haptic device for oral implantology,” Transactions on Edutainment VIII, vol. 12, pp. 232–240, 2012.  A. Gordon, I. Kim, A. C. Barnett, and J. Z. Moore, “Needle insertion force model for haptic simulation,” in Volume 2: Materials; Biomanufacturing; Properties, Applications and Systems; Sustainable Manufacturing, p. V002T03A003, Charlotte, North Carolina, USA, June, 2015.  J. Stern, I. S. Zeltser, and M. S. Pearle, “Percutaneous renal access simulators,” Journal of Endourology, vol. 21, no. 3, pp. 270–273, 2007.  Y. A. Noureldin and S. Andonian, “Simulation for percutane- ous renal access: where are we?,” Journal of Endourology, vol. 31, no. S1, pp. S-10–S-19, 2017. International Journal of Advances in Rotating Machinery Multimedia Journal of The Scientific Journal of Engineering World Journal Sensors Hindawi Hindawi Publishing Corporation Hindawi Hindawi Hindawi Hindawi www.hindawi.com Volume 2018 http://www www.hindawi.com .hindawi.com V Volume 2018 olume 2013 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 Journal of Control Science and Engineering Advances in Civil Engineering Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 Submit your manuscripts at www.hindawi.com Journal of Journal of Electrical and Computer Robotics Engineering Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 VLSI Design Advances in OptoElectronics International Journal of Modelling & Aerospace International Journal of Simulation Navigation and in Engineering Engineering Observation Hindawi Hindawi Hindawi Hindawi Volume 2018 Volume 2018 Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com www.hindawi.com www.hindawi.com Volume 2018 International Journal of Active and Passive International Journal of Antennas and Advances in Chemical Engineering Propagation Electronic Components Shock and Vibration Acoustics and Vibration Hindawi Hindawi Hindawi Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018
Applied Bionics and Biomechanics – Hindawi Publishing Corporation
Published: Jun 25, 2019
Access the full text.
Sign up today, get DeepDyve free for 14 days.