ABSTRACT. Accurately detecting and diagnosing developmental dysplasia of the hip (DDH), a common hip instability condition among infants, requires ultrasound (US) image data that captures the relevant anatomical structures. Our group has recently introduced a technique for automatically processing 3D US scans of the neonatal hip that significantly reduces dysplasia metric measurement variability, but it can be challenging for a clinician to know at the time of acquisition if they have acquired a volume suitable for analysis. We have previously introduced a single-slice-based approach for assessing adequacy of acquired volumes that worked reasonably well, but we believe that more explicitly three-dimensional approaches would be more robust and reliable. Here, we propose a new technique based on a convolutional neural network (CNN) architecture that incorporates inter-slice information and transfer learning. Our classifier labels volumes as adequate or inadequate for subsequent interpretation based on detecting the presence of key hip anatomical structures needed for DDH diagnosis. We validate our approach on 40 datasets from 15 pediatric patients and demonstrate a slice classification rate of 93% (improving on our previous implementation by 3%) with average processing time of 2 seconds per US volume. We expect automatic US scan adequacy assessment to have significant clinical impact with the potential to help in imaging standardization, improving efficiency of measuring DDH metrics, and improving accuracy of clinical decision making.

ABSTRACT. Many infants are born with unstable hips which can cause severe mobility issues as they grow older. This condition is known as developmental dysplasia of the hip (DDH), and although it is the most common hip disorder in infants, its diagnosis using 2D ultrasound (US) is prone to significant variability. This paper presents methods to reliably and automatically diagnose DDH in infants using 3D US. Based on 40 infant hip examinations, we demonstrate a 70% reduction in variability of 3D US-based measurements compared to those of standard 2D US-based measurements.

ABSTRACT. Abstract In total hip arthroplasty (THA), pelvic sagittal inclination (PSI) angle in the standing position is an important factor in planning of cup alignment. Previous methods for PSI estimation used a patient-specific CT to create digitally reconstructed radiographs (DRRs) and compare them with the radiograph to estimate PSI. In this study, we developed a method that estimates PSI from a single anteroposterior radiograph using two convolutional neural networks (CNNs). It does not require the patient-specific CT, thus reduces radiation exposure of the patient and opens up the possibility of application in a larger number of hospitals where CT is not acquired in a routine protocol. Our approach consists of two CNNs: one for the segmentation of the pelvis region in the radiograph and the other for the PSI estimation. For training of the networks, we used a database of 472 CTs acquired for the planning of THA. The pelvis region in each CT was automatically segmented with a previously developed method and the landmarks defining PSI were identified by an expert surgeon. First, we trained the segmentation network using many DRRs created from each CT with a random transformation and the corresponding pelvis mask. Then, the PSI estimation network is trained by DRRs and the corresponding PSI values. In this study, we evaluated performance of the segmentation network and the PSI estimation network separately using simulated radiographs (DRRs). The experiments demonstrated that the Dice coefficient of the segmentation network was 0.947±0.019. The PSI estimation error for all cases were 3.22±2.18 [deg]. We found that the accuracy was heavily dependent on the shape of the pelvis, because for 75% of the patients the PSI estimation error was relatively small (2.24±0.78 [deg]) for any PSI angles, while the other 25% resulted in a larger error (6.14±2.38 [deg]). Our future work includes tuning of the network architecture and the hyper parameters to improve estimation performance. Estimation of the pelvis shape (e.g., location of the anatomical landmarks in 3D) from radiograph is also underway.

ABSTRACT. In THA, femoral stem anteversion affects the range of motion until neck impingement on the cup rim. Although several combined anteversion theories have been proposed to minimize the risk of impingement, it has been found that adjustment with cup anteversion is not enough. In cases of DDH who often have high femoral neck anteversion, it has been reported that femoral neck anteversion is reduced functionally. To avoid the postoperative excessive functional internal rotation of the femur, reduction of the stem anteversion with a version changeable dual taper modular neck system may prevent unwanted postoperative functional internal rotation of the femur. The purpose of this study is to investigate whether it is possible to adjust the internal rotation of the hip joint using the neck system. We reviewed 108 hips of primary THA for osteoarthritis of females using a system with version changeable modular dual taper necks. We measured the anatomical anteversion and femoral rotational position on preoperative and postoperative CT images using the Kyocera 3D template software. Femoral anteversion was preoperatively average 22.0° in the straight neck group and 43.8° in the 15° reduced neck group. The change of femoral rotation angle was -5.6° and -3.9° in each neck groups. There was no significant difference in the change of femoral rotational position between the groups. This suggested that a postoperative excessive internal rotation of the femur in cases with high anteversion could be avoided by reducing femoral anteversion with the use of a retroverted modular neck as expected.

ABSTRACT. The purpose of the current study was to assess the relationship between occult periprosthetic acetabular fracture and bone mineral density (BMD) around the acetabulum using quantitative computed tomography (CT). The subjects were 64 patients who underwent primary total hip arthroplasty (THA) for osteoarthritis. They all had preoperative and postoperative three-dimensional CT scan with a bone density phantom. They were 59 females and 5 males. Their mean age was 63 years. The design of acetabular components was peripheral self-locking type in 48 hips and hemispherical type in 16 hips. BMD of the following 5 locations around the acetabulum were measured on quantitative CT; acetabular roof, anterior column and posterior column. Posterior column was divided into three parts, or upper, middle and lower parts. The occult fractures of the acetabulum were judged by the consultation among three expert hip surgeons using postoperative CT. BMD were compared between fracture group and non-fracture group. The periprosthetic occult fractures of the acetabulum were observed in 11 hips (17%). There was no significant difference in age between the groups with and without fracture. BMD of the acetabular roof in fracture group was higher than in non-fracture group. There were no significant differences in BMD of the anterior column and posterior column between the groups. This suggested that regional osteosclerotic change due to osteoarthritis had a greater effect on intraoperative occult fracture of the acetabulum in THA than osteoporosis.

ABSTRACT. Hip resurfacing arthroplasty (HRA) is a technical demanding procedure because malalignment of acetabular and femoral components cause serious postoperative complications. To avoid these complications, we have performed HRA using two types of CT-based navigation system. It is, however, unclear whether the navigation accuracy reaches a high level enough to overcome these complications in HRA. The purpose of this study was to evaluate the accuracy and precision of HRA performed using the CT-based navigation systems. HRA were performed in 17 hips of 16 patients using the CT-based navigation systems through posterolateral approach. The target of cup alignment was set as 40° in radiographic inclination and 15° in radiographic anteversion. The alignment of the femoral component stem was aimed to be parallel to the medial cortex of the femoral neck on the oblique coronal image through the femoral neck axis. Postoperative CT were acquired for postoperative analysis. We measured deviation of the cup alignment from the plan, deviation of the stem position and alignment from the plan and presence of femoral neck notching. The cup anteversion were within 5° from the plan in all cases. The cup inclination was within 5° from the planned inclination in 14 of 17 cases and within 10° in all cases. The angular difference of the stem were within 5° from the plan in all cases. There was no femoral neck notching. There were no postoperative complications including component loosening, dislocation and femoral neck fracture. The accuracy of component placement in HRA was high enough to avoid early complication.

ABSTRACT. Various types of contracture are accompanied with hip osteoarthritis (HOA) and they induce pelvic obliquity which can lead to functional leg length discrepancy. The purposes of this study was to evaluate morphological features of HOA that related to contracture using CT images. Three hundred seventeen hips of 292 patients who underwent primary total hip arthroplasty (THA) were the subjects of this study. Preoperative AP radiographs of the bilateral hips acquired in maximum adduction and maximum abduction were used to quantify contracture. Measuring angle, we classified as abduction contracture (group Ab), adduction contracture (group Ad) and control group (group C). To analyze the predictors for contracture, classification of HOA, anatomical factors ,spinal factors, femoral anteversion and the position of the osteophytes (Capital drop: Cd, Double floor: Df) were measured. Twenty six hips (8.1%) were classified into group Ab and 29 hips (9.1%) were into group Ad. Crowe classification, leg length discrepancy, offset, pelvic lateral inclination, femoral anteversion showed a significant correlation. In a multivariate analysis, abduction angle positively correlated with Crowe Ⅲ, and negatively correlated with the location of Cd and pelvic lateral inclination. The abduction angle, a positive correlation was found between the pelvic lateral incline and LLD, negative correlation was observed in Crow III. Additionally, Cd tended to locate lower than the end of Df in the Ab group. 　In conclusion, shortening leg length and higher Crowe classification relates to adduction contracture of the hip while osteophyte morphology related to abduction contracture of the hip.

ABSTRACT. Purpose: In theory, placing the tip of the fibula at the centre of the load-bearing region of the femoral head, close to the subchondral plate, is essential for early treatment of osteonecrosis of the femoral head by free vascularised fibular grafting. We introduce an orthopaedic robot-assisted system to insert the guide pin and place the fibula precisely. Methods: From September 2016 to November 2017, 11 patients (14 hips) with femoral head osteonecrosis had undergone robot-assisted free fibular grafting. Eleven hips were in Ficat stage II, and three hips were in Ficat stage III. Surgical Procedure: First, the descending branch of the lateral circumflex femoral artery was exposed. The Tian JI robot system was used to plan the entry point and target of the guide pin, control the robotic arm, guide it to place a cannula in the optimal position, and insert and calibrate the cannula for the guide pin. The guide pin was inserted and scanned to confirm its spatial position. Next, a bone window was created and some cancellous bone was harvested nearby, a tunnel was made by a cannulated drill with a diameter about 10mm along the guide pin, and the necrotic part of the femoral head was removed as totally as possible. The ipsilateral vascularized fibula was harvested, and the cancellous bone and/or artificial bone was used to fill the cavity after grinding the necrotic area of the bone. Finally, the vessels were anastomosised. Conventional anticoagulant, anti-infective therapy was performed after the procedure. Results: All 14 surgical procedures were successful. The guide pins and fibulae were accurately placed according to the robot’s plan, and the tips of the fibulae were placed at the centre of the load-bearing region of the femoral heads, 4 to 6 mm from the articular surface. The patients were followed up postoperatively for three to fifteen months, with an average of seven months. The function of the hip joint recovered smoothly for 10 patients. Frontal and lateral X-ray and CT scans showed that the tips of the fibulae were placed at the centre of the load-bearing region, 4 to 6 mm from the articular surface. Conclusion: With the assistance of an orthopaedic robot system, the guide pin can be accurately positioned, thereby allowing the tip of the fibula to be inserted into the optimal anatomical position and maximising its mechanical efficacy. The clinical effects require more cases, more rigorous studies, and longer-term follow-up to be validated.

ABSTRACT. Objective: To investigate an effective method to reconstruct periacetabular defect using a novel 3D printing individualized prosthesis after accurate en bloc tumor resection with the help of computer assisted navigation system (CANS) or osteotomy guide plate. Methods: 3D model of image fusion was used in preoperative design. A novel 3D printing individualized acetabular prosthesis was designed and manufactured using 3D metal printing techniques (EBM). The tumor was excised accurately and the prosthesis was fixed precisely by CAN or guide plate. From December 2010 to May 2016, 8 patients with periacetabular tumor (chondrosarcoma 5 cases and GCT 3 cases) were performed. The tumor control, function and complication were evaluated postoperatively. Results: 8 cases were treated by accurate en-bloc resection and precise reconstruction with a novel 3D printing individualized prosthesis．One case suffered dislocation of hip joint and two cases presented deep venous thrombosis. The mean follow-up period was 37.5 months (range, 5-70 months). 6 patients survived free of disease. One patient was subject to local recurrence after one year. One patient died from distal metastasis after three years. No infection, prosthesis loosening and breakage occurred. The MSTS score was 18~27．Conclusion: Accurate tumor resection and precise reconstruction with 3D printing individualized prosthesis can be achieved by CANS or guide plate. 3D printing implants might be a promising approach for precision reconstruction and long-term stability.

ABSTRACT. Screw fixation in upper cervical spine surgery is one of the most challenging spinal procedures. For that reason, the free-hand techniques are not a reliable procedure in upper cervical spine. Wei Tian reported the first posterior C1-2 transarticular screw fixation which was assisted by robotic systems using TiRobot system in 2016. Our study was aimed to assess the accuracy and reliability of screw fixation in upper cervical spine prospectively. All the patients undergoing screw fixation assisted by robotic system (TiRobot) in upper cervical spine were prospectively studied from August 2015 to January 2018 in Beijing Jishuitan hospital. During surgeries, intraoperatively obtained images by C-arm were transferred into the TiRobot system and three-dimensional images were created. Surgeons' plannings of the screw trajectories were performed in TiRobot system. Afterwards, the robot arm with a guidance tube on its end was automatically moved to the entry point of each trajectory and held still. Guiding pins were inserted and followed by cannulated or conventional screws placements. A fluoroscopic re-scan by C-arm was performed, and the merging of two sets of images demonstrated the deviations between the planned and real trajectories. The deviations were analyzed and any intraoperative difficulties or errors were recorded. Postoperatively, any screw perforations were reviewed and recorded on CT scan images. Twenty-six patients (13 males/13 females) were included in this study. The average age was forty-nine. The underlying diseases of these patients covered dens fractures, Hangman's fractures, congenial and acquired deformities. 64 screws were placed assisted by TiRobot system. The average deviation between the planned and real trajectories was 0.9 mm. No screw perforations were detected on postoperative CT scans. Cervical spine, particularly the upper cervical spine has small size bony structures, higher accuracy requirements are significantly important. The results of our study showed only 0.9 mm deviations between planned and real trajectories occurred. This level of accuracy allowed safe and accurate placement of screws in upper cervical spine and no screw perforations were observed. For that reasons, screw fixation in upper cervical spine is an accurate and reliable procedure using TiRobot system.

ABSTRACT. The orientation of the cup in Total Hip Arthroplasty plays a major role on the post-operative results. It has been considered for several decades that the cup should be oriented according to the safe zone defined by Lewinnek. However, this safe zone is not always suitable because of the inter-individual variabilities of the pelvic tilt during daily activities. We propose in this paper a non-invasive ultrasound based solution which can easily measure this patient specific parameter in order to thereafter otimise the cup orientation. The accuracy of this system was assessed with a specific pelvic phantom. A clinical pilot study was also performed on ten patients. The pelvic tilt was measured in three daily positions: the supine, sitting and standing positions. The average error was 1.15°±0.82°. The average pelvic tilt was -97.1°±28.6°, -46.3°±12.8° and -9.0°±8.3° for respectively the supine, the sitting and the standing positions. The high inter-individual variabilities of the pelvic tilt in different daily positions highlight the need to have a suitable device for the measurement of this patient specific parameter for THA. The proposed system is easy-to-use, portable and allows the pelvic tilt measurement in different positions without any additional x-ray radiation.

ABSTRACT. LEARNING CURVE IN COMPUTER-ASSISTED TOTAL KNEE ARTHROPLASTY: A CUSUM ANALYSIS Y. Dai*, G. Bras and C. Hamad *Exactech Inc, Gainesville, FL, 32653, USA, Yifei.dai@exac.com

Limited studies have assessed the learning of computer-assisted orthopaedic system (CAOS) in total knee arthroplasty (TKA). This may due to the limitation of currently methodologies, which was challenged by the noisy nature of the surgical time data. This study adopted an advanced methodology (CUSUM) to analyse the learning of a CAOS system. The first 50 CAOS TKAs from 10 surgeons (7 senior/3 novice) were analysed. CUSUM of deviances in surgical time was analysed for each surgeon. A horizontal trend in the chronological CUSUM plot signified the stabilization of the process. The first case (cases to proficiency: CP) of the horizontal trend marked the end of learning for each surgeon. The CP was compared between senior and novice surgeons. The difference in surgical time was compared between cases during (#1-CP) and after (#41-50) learning. Each surgeon’s CUSUM plot demonstrated the completion of learning by exhibiting the stabilization trend. On average, it took 12-13 cases to complete learning with no substantial difference between senior and novice surgeons. Both surgeon groups spent ~15min more in surgery during learning than their cases #41-50, with novice surgeons having ~3min more in time increase compared to senior surgeons. The data suggested that the average learning of the CAOS system takes 12-13 cases, disregard surgeon’s experience. The learning phase moderately increased surgical time, without substantial difference between senior and novice surgeons. Having no CAOS experience did not result in substantially steeper learning curve. Use of advanced method in studying learning curve provided improved understanding of CAOS learning.

ABSTRACT. Symptomatic malunions of the distal radius are often treated with correction osteotomies. However, complex bone deformities can often be difficult to identify as these measures are not completely addressed by conventional preoperative planning techniques. From January 2014 to January 2018 23 patients with symptomatic distal radius malunions were treated in our hospital using a computer-assisted preoperative planning technique. We compared these patients with a control group that consisted of 23 patients with similar deformities that underwent surgery with a conventional planning approach. The computer-assisted group underwent a CT scan of the injured forearm as well as the contralateral extremity preoperatively. After that 3D anatomical models were created using specialized segmentation software. The planning process was carried out in a computer program developed by our team. A patient-specific surgical guide was then produced on a 3D printer. In both groups volar locking plates of the same manufacturer were used. We examined the postoperative x-rays, range of motion (ROM) and Disabilities of the Arm, Shoulder and Hand (DASH) score at 3, 6 and 12 months postoperatively in both groups. At 3 months postoperatively all patients in the computer-assisted group showed recovery of ROM, with no difference compared to the contralateral side, 6 patients in the conventional planning group had reduced ROM. Postoperative x-rays revealed a residual volar tilt in these patients. The computer-assisted group had no anatomical abnormalities on x-ray. At 12 months postoperatively there were no significant differences in ROM in both groups. Patients in both groups had total pain relief. Computer-assisted planning with the use of 3D printed patient-specific surgical guides enhances results of corrective osteotomies of distal radius malunions, facilitating transfer of the computer generated plan to the operating room.

ABSTRACT. The bone fragment and fracture lines may not clear enough to be visualized after bone reduction and may be hidden behind other bones. A 3D printing technology can be used to deal with this drawbacks. A series of algorithms were developed to ensure that two adjacent bone fragments fabricated can be assembled successfully. In this study, we developed a bone reduction algorithm, and a mesh-overlapping detection and removal algorithm to ensure that two 3D-printed fragments are easily assembled and disassembled. In addition, we also developed a method to detect and extract fracture boundary from two adjacent fragments, and then apply a multi-color 3D printing technology to fabricate the assembled bone fragments. The advantage of the proposed method is that we can show the status of 3D-printed bone fragments before and after bone reduction, with fracture lines printed in different color.

ABSTRACT. We report a patient who was successfully treated with minimally invasive tubular spondylectomy for C4 metastasis tumor. pathological diagnosis was thyroid follicular carcinoma.Surgical indication was definite because there was a risk of spinal cord injury as a result of vertebral destruction and tumor extension into the epidural space.We didn’t search any relevant literature about minimally invasive tubular spondylectomy.We peformed one-stage minimally invasive tubular spondylectomy by anterior and posterior approaches.Operative procedures:The anterior aspect of the spine was exposed from C3 to C5 by tubular retractors combined with common retractors.a longitudinal skin incision about 4cm.Obtained a marginal resection.The 3D printing artificial vertebral body was fitted into the C4 vertebral body site and checked. An anterior plate was fixed.The procedure was performed through a midline posterior cervical incision in line with the c3-5 spinous processes. The incision was 3 cm in length. Ultrasound osteotome and osteotome used to cut posterior element to about four to five blocks,then take out one by one.C3-5 performed navigation assisted lateral mass screw fixation.Operation is successful without complication.Pathological diagnosis was metastatic thyroid carcinoma.Postoperative VAS Scores:0.American Spinal Injury Association grade E at the latest follow-up.There were no local recurrences in 3 months follow-up.Indication and contraindication need to do further study.We combined application of tubular retractors and common retractors in anterior operation.Tubular retractors can finish posterior operation through cutting bone to segments.3D printing artificial vertebral body provided better stability and less subsidence than titanium mesh.When working through tubular retraction systems, the inability to obtain adequate visualization can be the major difficulty of the procedure.Surgeon must avoid risk of vertebral arteries injury, improper instrumentation placement, or neurologic injury due to poor access and visualization.:Minimally invasive tubular spondylectomy is feasible.But the operation is very complicated and high risk.Surgeons must have sufficient experience about both open operation and minimally invasive tubular operation.Learning curve is cliff and long.

ABSTRACT. Purpose : A new algorithm, based on Fully Convolutional Networks (FCN), is proposed for the automatic localization of the bone interface in ultrasound (US) images. The aim of this paper is to compare and validate this method with (1) a manual segmentation that was performed by three independent experts, and (2) a state of the art method called Confidence in Phase Symmetry (CPS).

Methods : The dataset used for this study is composed of 3692 US images collected from three volunteers on six different anatomical structures. Three experts manually delineated the bone interface on these US images. The inter- and intra observer variabilities of this manual delineation was assessed. Both FCN based and CPS approaches were studied and compared to the inter-observer average segmentation according to six criteria: recall, precision, F1-Score, accuracy, specificity and Root Mean Square Error (RMSE).

Results : The intra- and inter-observer variabilities were inferior to 0.8 mm for 90% of manual annotations. The RMSE was 0.99 ± 2.62 mm and 6.43 ± 7.66 mm for respectively the FCN based approach and the CPS algorithm. The mean recall, precision, f1-Score, accuracy and specificity were respectively 66%, 65%, 62%, 81% and 84% for the FCN based approach, and 64%, 36%, 41%, 50% and 41% for the CPS algorithm.

Conclusion : The FCN based approach outperforms the CPS algorithm. The obtained RMSE is close to the manual segmentation variability.

ABSTRACT. Objective: To explore the feasibility and accuracy of the new technique of patient-specific 3D printing screws insertion template in the pelvic fractures. Methods: From January to June in 2017, 6 patients with pelvic fractures were treated by this new type of screw guide templates. Pubis screws and sacroiliac screws were inserted. The patients with external fixation were examined by CT scan before surgery. The patients-specific 3D printing templates were made with photosensitive resin by a 3D printing system to ensure the trajectory of the screws. The templates were sterilized by ethylene oxide and used during surgery. The accuracy and safety of the templates were evaluated by CT scans after surgery. Results: Totally 10 screws were inserted, including 5 pubis screws and 5 sacroiliac screws. The average surgical time for pubis screw was 12.7 min/screw, and 9.2 min/screw for sacroiliac screw. The average time of X-ray exposure was 13.3±3.6s/screw for public screw and 9.6±4.5s/screw for sacroiliac screw. All the screws were inserted in the trajectory as preoperative design and the screw axis deviation was 1.60 ± 0.2mm and 2 ± 0.3 degrees angle deviation by preoperative and postoperative CT verification. Blood loss during the surgery was minimal (12.2ml/screw). Vascular of neurologic complications or injuries did not happen. And no infection, broken screws or screw pullout occurred. ConculsionThe patient-specific screw guide template based on the external fixation can insert the screw accurately and safely with very small incision. This technique is a new kind of intraoperative screw navigations. The patient-specific 3D printing screw insertion template was user-frindly, moderate cost and enabled a radiation-reduced pelvic screw insertion.

ABSTRACT. Percutaneous iliosacral screw fixation is a common treatment for sacral fracture to stabilize pelvic and acetabular fractures. This procedure requires substantial experience and anatomical knowledge to determine a safe entry point and the trajectory. Here we demonstrated a computer-aided method for determining an optimal and safe osseous fixation pathway (OFP) for percutaneous transverse iliosacral screw fixation in the first sacral segment. The first part of the study involved 50 healthy volunteers who received pelvic CT scans. We reconstructed the pelvic CT images, produced multi-slice sagittal views, and determined the axis for safe OFP. The second part included 18 patients with longitudinal sacral fractures. They received percutaneous transverse iliosacral screw, which was preoperatively planned by the method described here. The deviation between the actual and the planned screw position was measured, and the operation time and all complications were recorded. The mean operative time was 30.15±5.08 minutes. No patients encountered perioperative complications. The mean length of the optimal OFP was 135.30±7.80 mm and the mean length of the iliosacral screw was 115.16±4.18 mm. Postoperative pelvic X-ray and CT scan showed satisfactory screw positions. The mean deviation of the screws was 5.3° ± 2.2° in horizontal and 4.1°±2.4° in vertical. The deviation was within the safe zone. With the use of 3D CT reconstruction images, the method demonstrated here was applicable for preoperative planning a safe OFP for percutaneous transverse iliosacral screw in longitudinal sacral fractures.