Semi-Supervised Bladder Tissue Classification in Multi-Domain Endoscopic Images.

OBJECTIVE: Accurate visual classification of bladder tissue during Trans-Urethral Resection of Bladder Tumor (TURBT) procedures is essential to improve early cancer diagnosis and treatment. During TURBT interventions, White Light Imaging (WLI) and Narrow Band Imaging (NBI) techniques are used for lesion detection. Each imaging technique provides diverse visual information that allows clinicians to identify and classify cancerous lesions. Computer vision methods that use both imaging techniques could improve endoscopic diagnosis. We address the challenge of tissue classification when annotations are available only in one domain, in our case WLI, and the endoscopic images correspond to an unpaired dataset, i.e. there is no exact equivalent for every image in both NBI and WLI domains. METHOD: We propose a semi-surprised Generative Adversarial Network (GAN)-based method composed of three main components: a teacher network trained on the labeled WLI data; a cycle-consistency GAN to perform unpaired image-to-image translation, and a multi-input student network. To ensure the quality of the synthetic images generated by the proposed GAN we perform a detailed quantitative, and qualitative analysis with the help of specialists. CONCLUSION: The overall average classification accuracy, precision, and recall obtained with the proposed method for tissue classification are 0.90, 0.88, and 0.89 respectively, while the same metrics obtained in the unlabeled domain (NBI) are 0.92, 0.64, and 0.94 respectively. The quality of the generated images is reliable enough to deceive specialists. SIGNIFICANCE: This study shows the potential of using semi-supervised GAN-based bladder tissue classification when annotations are limited in multi-domain data.

Robotic Instruments Inside the MRI Bore: Key Concepts and Evolving Paradigms in Imaging-enhanced Cranial Neurosurgery.

Intraoperative MRI has been increasingly used to robotically deliver electrodes and catheters into the human brain using a linear trajectory with great clinical success. Current cranial MR guided robotics do not allow for continuous real-time imaging during the procedure because most surgical instruments are not MR-conditional. MRI guided robotic cranial surgery can achieve its full potential if all the traditional advantages of robotics (such as tremor-filtering, precision motion scaling, etc.) can be incorporated with the neurosurgeon physically present in the MRI bore or working remotely through controlled robotic arms. The technological limitations of design optimization, choice of sensing, kinematic modeling, physical constraints, and real-time control had hampered early developments in this emerging field, but continued research and development in these areas over time has granted neurosurgeons far greater confidence in using cranial robotic techniques. This article elucidates the role of MR-guided robotic procedures using clinical devices like NeuroBlate and Clearpoint that have several thousands of cases operated in a "linear cranial trajectory" and planned clinical trials, such as LAANTERN for MR guided robotics in cranial neurosurgery using LITT and MR-guided putaminal delivery of AAV2 GDNF in Parkinson's disease. The next logical improvisation would be a steerable curvilinear trajectory in cranial robotics with added DOFs and distal tip dexterity to the neurosurgical tools. Similarly, the novel concept of robotic actuators that are powered, imaged, and controlled by the MRI itself is discussed in this article, with its potential for seamless cranial neurosurgery.

FUN-SIS: A Fully UNsupervised approach for Surgical Instrument Segmentation.

Automatic surgical instrument segmentation of endoscopic images is a crucial building block of many computer-assistance applications for minimally invasive surgery. So far, state-of-the-art approaches completely rely on the availability of a ground-truth supervision signal, obtained via manual annotation, thus expensive to collect at large scale. In this paper, we present FUN-SIS, a Fully-UNsupervised approach for binary Surgical Instrument Segmentation. FUN-SIS trains a per-frame segmentation model on completely unlabelled endoscopic videos, by solely relying on implicit motion information and instrument shape-priors. We define shape-priors as realistic segmentation masks of the instruments, not necessarily coming from the same dataset/domain as the videos. The shape-priors can be collected in various and convenient ways, such as recycling existing annotations from other datasets. We leverage them as part of a novel generative-adversarial approach, allowing to perform unsupervised instrument segmentation of optical-flow images during training. We then use the obtained instrument masks as pseudo-labels in order to train a per-frame segmentation model; to this aim, we develop a learning-from-noisy-labels architecture, designed to extract a clean supervision signal from these pseudo-labels, leveraging their peculiar noise properties. We validate the proposed contributions on three surgical datasets, including the MICCAI 2017 EndoVis Robotic Instrument Segmentation Challenge dataset. The obtained fully-unsupervised results for surgical instrument segmentation are almost on par with the ones of fully-supervised state-of-the-art approaches. This suggests the tremendous potential of the proposed method to leverage the great amount of unlabelled data produced in the context of minimally invasive surgery.

Distortion and instability compensation with deep learning for rotational scanning endoscopic optical coherence tomography.

Optical Coherence Tomography (OCT) is increasingly used in endoluminal procedures since it provides high-speed and high resolution imaging. Distortion and instability of images obtained with a proximal scanning endoscopic OCT system are significant due to the motor rotation irregularity, the friction between the rotating probe and outer sheath and synchronization issues. On-line compensation of artefacts is essential to ensure image quality suitable for real-time assistance during diagnosis or minimally invasive treatment. In this paper, we propose a new online correction method to tackle both B-scan distortion, video stream shaking and drift problem of endoscopic OCT linked to A-line level image shifting. The proposed computational approach for OCT scanning video correction integrates a Convolutional Neural Network (CNN) to improve the estimation of azimuthal shifting of each A-line. To suppress the accumulative error of integral estimation we also introduce another CNN branch to estimate a dynamic overall orientation angle. We train the network with semi-synthetic OCT videos by intentionally adding rotational distortion into real OCT scanning images. The results show that networks trained on this semi-synthetic data generalize to stabilize real OCT videos, and the algorithm efficacy is demonstrated on both ex vivo and in vivo data, where strong scanning artifacts are successfully corrected.

Shape Reconstruction Processes for Interventional Application Devices: State of the Art, Progress, and Future Directions.

Soft and continuum robots are transforming medical interventions thanks to their flexibility, miniaturization, and multidirectional movement abilities. Although flexibility enables reaching targets in unstructured and dynamic environments, it also creates challenges for control, especially due to interactions with the anatomy. Thus, in recent years lots of efforts have been devoted for the development of shape reconstruction methods, with the advancement of different kinematic models, sensors, and imaging techniques. These methods can increase the performance of the control action as well as provide the tip position of robotic manipulators relative to the anatomy. Each method, however, has its advantages and disadvantages and can be worthwhile in different situations. For example, electromagnetic (EM) and Fiber Bragg Grating (FBG) sensor-based shape reconstruction methods can be used in small-scale robots due to their advantages thanks to miniaturization, fast response, and high sensitivity. Yet, the problem of electromagnetic interference in the case of EM sensors, and poor response to high strains in the case of FBG sensors need to be considered. To help the reader make a suitable choice, this paper presents a review of recent progress on shape reconstruction methods, based on a systematic literature search, excluding pure kinematic models. Methods are classified into two categories. First, sensor-based techniques are presented that discuss the use of various sensors such as FBG, EM, and passive stretchable sensors for reconstructing the shape of the robots. Second, imaging-based methods are discussed that utilize images from different imaging systems such as fluoroscopy, endoscopy cameras, and ultrasound for the shape reconstruction process. The applicability, benefits, and limitations of each method are discussed. Finally, the paper draws some future promising directions for the enhancement of the shape reconstruction methods by discussing open questions and alternative methods.

Using spatial-temporal ensembles of convolutional neural networks for lumen segmentation in ureteroscopy.

PURPOSE: Ureteroscopy is an efficient endoscopic minimally invasive technique for the diagnosis and treatment of upper tract urothelial carcinoma. During ureteroscopy, the automatic segmentation of the hollow lumen is of primary importance, since it indicates the path that the endoscope should follow. In order to obtain an accurate segmentation of the hollow lumen, this paper presents an automatic method based on convolutional neural networks (CNNs). METHODS: The proposed method is based on an ensemble of 4 parallel CNNs to simultaneously process single and multi-frame information. Of these, two architectures are taken as core-models, namely U-Net based in residual blocks ([Formula: see text]) and Mask-RCNN ([Formula: see text]), which are fed with single still-frames I(t). The other two models ([Formula: see text], [Formula: see text]) are modifications of the former ones consisting on the addition of a stage which makes use of 3D convolutions to process temporal information. [Formula: see text], [Formula: see text] are fed with triplets of frames ([Formula: see text], I(t), [Formula: see text]) to produce the segmentation for I(t). RESULTS: The proposed method was evaluated using a custom dataset of 11 videos (2673 frames) which were collected and manually annotated from 6 patients. We obtain a Dice similarity coefficient of 0.80, outperforming previous state-of-the-art methods. CONCLUSION: The obtained results show that spatial-temporal information can be effectively exploited by the ensemble model to improve hollow lumen segmentation in ureteroscopic images. The method is effective also in the presence of poor visibility, occasional bleeding, or specular reflections.

Minimally Invasive Bilateral Anterior Cingulotomy via Open Minicraniotomy Using a Novel Multiport Cisternoscope: A Cadaveric Demonstration.

BACKGROUND: Bilateral anterior cingulotomy has been used to treat chronic pain, obsessive compulsive disorder, and addictions. Lesioning of the target area is typically performed using bilateral stereotactic electrode placement and target ablation, which involves transparenchymal access through both hemispheres. OBJECTIVE: To evaluate an endoscopic direct-vision lesioning using a unilateral parasagittal minicraniotomy for minimally invasive bilateral anterior cingulotomy using a novel multiport endoscope through the anterior interhemispheric fissure. METHODS: A novel multiport magnetic resonance imaging (MRI)-compatible neuroendoscope prototype is used to demonstrate cadaveric cingulate lesioning through a lateral imaging port while simultaneously viewing the pericallosal arteries as landmarks through a tip imaging port. The lateral port enables extended lesioning of the gyrus while rotation of the endoscope about its axis provides access to homologous areas of both hemispheres. RESULTS: Cadaver testing confirmed the capability to navigate the multiport neuroendoscope between the hemispheres using concurrent imaging from the tip and lateral ports. The lateral port enabled exploration of the gyrus, visualization of lesioning, and subsequent inspection of lesions. Tip-port imaging provided navigational cues and allowed the operator to ensure that the endoscope tip did not contact tissue. The multiport design required instrument rotation in the coronal plane of only 20° to lesion both gyri, while a standard endoscope necessitated a rotation of 54°. CONCLUSION: Multiport MRI-compatible endoscopy can be effectively used in cisternal endoscopy, whereby a unilateral parasagittal minicraniotomy can be used for endoscopic interhemispheric bilateral anterior cingulotomy.

Combining Differential Kinematics and Optical Flow for Automatic Labeling of Continuum Robots in Minimally Invasive Surgery.

The segmentation of continuum robots in medical images can be of interest for analyzing surgical procedures or for controlling them. However, the automatic segmentation of continuous and flexible shapes is not an easy task. On one hand conventional approaches are not adapted to the specificities of these instruments, such as imprecise kinematic models, and on the other hand techniques based on deep-learning showed interesting capabilities but need many manually labeled images. In this article we propose a novel approach for segmenting continuum robots on endoscopic images, which requires no prior on the instrument visual appearance and no manual annotation of images. The method relies on the use of the combination of kinematic models and differential kinematic models of the robot and the analysis of optical flow in the images. A cost function aggregating information from the acquired image, from optical flow and from robot encoders is optimized using particle swarm optimization and provides estimated parameters of the pose of the continuum instrument and a mask defining the instrument in the image. In addition a temporal consistency is assessed in order to improve stochastic optimization and reject outliers. The proposed approach has been tested for the robotic instruments of a flexible endoscopy platform both for benchtop acquisitions and an in vivo video. The results show the ability of the technique to correctly segment the instruments without a prior, and in challenging conditions. The obtained segmentation can be used for several applications, for instance for providing automatic labels for machine learning techniques.

A low-cost bioprosthetic semilunar valve for research, disease modelling and surgical training applications.

OBJECTIVES: This paper provides detailed instructions for constructing low-cost bioprosthetic semilunar valves for animal research and clinical training. This work fills an important gap between existing simulator training valves and clinical valves by providing fully functioning designs that can be employed in ex vivo and in vivo experiments and can also be modified to model valvular disease. METHODS: Valves are constructed in 4 steps consisting of creating a metal frame, covering it with fabric and attaching a suture ring and leaflets. Computer-aided design files are provided for making the frame from wire or by metal 3D printing. The covering fabric and suturing ring are made from materials readily available in a surgical lab, while the leaflets are made from pericardium. The entire fabrication process is described in figures and in a video. To demonstrate disease modelling, design modifications are described for producing paravalvular leaks, and these valves were evaluated in porcine ex vivo (n = 3) and in vivo (n = 6) experiments. RESULTS: Porcine ex vivo and acute in vivo experiments demonstrate that the valves can replicate the performance of clinical valves for research and training purposes. Surgical implantation is similar, and echocardiograms are comparable to clinical valves. Furthermore, valve leaflet function was satisfactory during acute in vivo tests with little central regurgitation, while the paravalvular leak modifications consistently produced leaks in the desired locations. CONCLUSIONS: The detailed design procedure presented here, which includes a tutorial video and computer-aided design files, should be of substantial benefit to researchers developing valve disease models and to clinicians developing realistic valve training systems.

Cardioscopically Guided Beating Heart Surgery: Paravalvular Leak Repair.

PURPOSE: There remains a paucity of direct visualization techniques for beating-heart intracardiac procedures. To address this need, we evaluated a novel cardioscope in the context of aortic paravalvular leaks (PVLs) localization and closure. DESCRIPTION: A porcine aortic PVL model was created using a custom-made bioprosthetic valve, and PVL presence was verified by epicardial echocardiography. Transapical delivery of occlusion devices guided solely by cardioscopy was attempted 13 times in a total of three pigs. Device retrieval after release was attempted six times. Echocardiography, morphologic evaluation, and delivery time were used to assess results. EVALUATION: Cardioscopic imaging enabled localization of PVLs via visualization of regurgitant jet flow in a paravalvular channel at the base of the prosthetic aortic valve. Occluders were successfully placed in 11 of 13 attempts (84.6%), taking on average 3:03 ± 1:34 min. Devices were cardioscopically removed successfully in three of six attempts (50%), taking 3:41 ± 1:46 min. No damage to the ventricle or annulus was observed at necropsy. CONCLUSIONS: Cardioscopy can facilitate intracardiac interventions by providing direct visualization of anatomic structures inside the blood-filled, beating-heart model.

A multiport MR-compatible neuroendoscope: spanning the gap between rigid and flexible scopes.

OBJECTIVE Rigid endoscopes enable minimally invasive access to the ventricular system; however, the operative field is limited to the instrument tip, necessitating rotation of the entire instrument and causing consequent tissue compression while reaching around corners. Although flexible endoscopes offer tip steerability to address this limitation, they are more difficult to control and provide fewer and smaller working channels. A middle ground between these instruments-a rigid endoscope that possesses multiple instrument ports (for example, one at the tip and one on the side)-is proposed in this article, and a prototype device is evaluated in the context of a third ventricular colloid cyst resection combined with septostomy. METHODS A prototype neuroendoscope was designed and fabricated to include 2 optical ports, one located at the instrument tip and one located laterally. Each optical port includes its own complementary metal-oxide semiconductor (CMOS) chip camera, light-emitting diode (LED) illumination, and working channels. The tip port incorporates a clear silicone optical window that provides 2 additional features. First, for enhanced safety during tool insertion, instruments can be initially seen inside the window before they extend from the scope tip. Second, the compliant tip can be pressed against tissue to enable visualization even in a blood-filled field. These capabilities were tested in fresh porcine brains. The image quality of the multiport endoscope was evaluated using test targets positioned at clinically relevant distances from each imaging port, comparing it with those of clinical rigid and flexible neuroendoscopes. Human cadaver testing was used to demonstrate third ventricular colloid cyst phantom resection through the tip port and a septostomy performed through the lateral port. To extend its utility in the treatment of periventricular tumors using MR-guided laser therapy, the device was designed to be MR compatible. Its functionality and compatibility inside a 3-T clinical scanner were also tested in a brain from a freshly euthanized female pig. RESULTS Testing in porcine brains confirmed the multiport endoscope's ability to visualize tissue in a blood-filled field and to operate inside a 3-T MRI scanner. Cadaver testing confirmed the device's utility in operating through both of its ports and performing combined third ventricular colloid cyst resection and septostomy with an endoscope rotation of less than 5°. CONCLUSIONS The proposed design provides freedom in selecting both the number and orientation of imaging and instrument ports, which can be customized for each ventricular pathological entity. The lightweight, easily manipulated device can provide added steerability while reducing the potential for the serious brain distortion that happens with rigid endoscope navigation. This capability would be particularly valuable in treating hydrocephalus, both primary and secondary (due to tumors, cysts, and so forth). Magnetic resonance compatibility can aid in endoscope-assisted ventricular aqueductal plasty and stenting, the management of multiloculated complex hydrocephalus, and postinflammatory hydrocephalus in which scarring obscures the ventricular anatomy.

Step-and-Repeat Nanoimprint-, Photo- and Laser Lithography from One Customised CNC Machine.

The conversion of a computer numerical control machine into a nanoimprint step-and-repeat tool with additional laser- and photolithography capacity is documented here. All three processes, each demonstrated on a variety of photoresists, are performed successfully and analysed so as to enable the reader to relate their known lithography process(es) to the findings. Using the converted tool, 1 cm(2) of nanopattern may be exposed in 6 s, over 3300 times faster than the electron beam equivalent. Nanoimprint tools are commercially available, but these can cost around 1000 times more than this customised computer numerical control (CNC) machine. The converted equipment facilitates rapid production and large area micro- and nanoscale research on small grants, ultimately enabling faster and more diverse growth in this field of science. In comparison to commercial tools, this converted CNC also boasts capacity to handle larger substrates, temperature control and active force control, up to ten times more curing dose and compactness. Actual devices are fabricated using the machine including an expanded nanotopographic array and microfluidic PDMS Y-channel mixers.

Fluorinated ethylene-propylene: a complementary alternative to PDMS for nanoimprint stamps.

Polydimethylsiloxane (PDMS) is used by many for nanoimprint applications due to its affordability, ease of preparation, mechanical flexibility, compatibility with imprint resists and transparency to UV light. However PDMS is notoriously flexible, tacky and permeable to air. Here fluorinated ethylene-propylene (FEP) is considered as a viable and versatile alternative material for nanoimprint stamps. FEP possesses many of the desirable nanoimprint attributes associated with PDMS but crucially also features a range of complementary characteristics, including an order of magnitude more mechanical strength allowing it to handle higher loads than PDMS, an intrinsically non-stick surface and is compatible with oxygen sensitive resists. Unlike elastomeric polymers, FEP is glassy so patterning may be realised via hot embossing. Not only is this a facile and rapid means of physical structuring but it also facilitates combinatorial patterning, providing a versatility beyond that of traditional casting materials. Due to the intrinsically slow creep of FEP both micro- and nanopatterning are successfully performed sequentially. Feature sizes from 45 nm were successfully realised via the hot-embossing method. To further demonstrate the potential of the material, a modified computer numerical control machine is used. It is capable of photo-, nanoimprint- and laser lithography in conjunction with patterned FEP foils. The tool is used to perform pattern transfer into a developmental nanoimprint resist from Micro Resist Technology, mr-NIL210 XP, and Nano SU-8 3005 negative tone photo resist from MicroChem. Ultimately three-tier lithography is performed in unison and advantageous step-and-repeat performance is achieved with fabricated FEP imprint stamps as they demould more compliantly and resist pressure and contamination better than PDMS.

Control of a hybrid robotic system for computer-assisted interventions in dynamic environments.

PURPOSE: Minimally invasive surgery is becoming the standard treatment of care for a variety of procedures. Surgeons need to display a high level of proficiency to overcome the challenges imposed by the minimal access. Especially when operating on a dynamic organ, it becomes very difficult to align instruments reliably and precisely. In this paper, a hybrid rigid/continuum robotic system and a dedicated robotic control approach are proposed to assist the surgeon performing complex surgical gestures in a dynamic environment. METHODS: The proposed robotic system consists of a rigid robot arm on top of which a continuum robot is mounted in series. The continuum robot is locally actuated with McKibben muscles. A control scheme based on quadratic programming framework is adopted. It is shown that the framework allows enforcing a set of constraints on the pose of the tip, as well as of the instrument shaft, which is commanded to slide in and out through the entry point. RESULTS: Through simulation and experiments, it is shown how the robot tool tip is able to follow sinusoidal trajectories of 0.37 and 2 Hz, while maintaining the instrument shaft pivoting along the entry point. The positioning and tracking accuracy of such system are shown to lie below 4.7 mm in position and [Formula: see text] in angle. CONCLUSION: The results suggest a good potential for applying the proposed technology to assist the surgeon during complex robot-assisted interventions. It is also illustrated that even when using flexible hence relatively safe end-effectors, it is possible to reach acceptable tracking behaviour at relatively high frequencies.

Estimation of optimal pivot point for remote center of motion alignment in surgery.

PURPOSE: The determination of an optimal pivot point ([Formula: see text]) is important for instrument manipulation in minimally invasive surgery. Such knowledge is of particular importance for robotic-assisted surgery where robots need to rotate precisely around a specific point in space in order to minimize trauma to the body wall while maintaining position control. Remote center of motion (RCM) mechanisms are commonly used, where the RCM point is manually and visually aligned. If not positioned appropriately, this misalignment might lead to intolerably high forces on the body wall with increased risk of postoperative complications or instrument damage. An automated method to align the RCM with the [Formula: see text] was developed and tested. METHOD: Computer vision and a lightweight calibration procedure are used to estimate the optimal pivot point. One or two pre-calibrated cameras viewing the surgical scene are employed. The surgeon is asked to make short pivoting movements, applying as little torque as possible, with an instrument of choice passing through the insertion point while camera images are being recorded. The physical properties of an instrument rotating around a pivot point are exploited in a random sample consensus scheme to robustly estimate the ideal position of the RCM in the image planes. Triangulation is used to estimate the RCM position in 3D. Experiments were performed on a specially designed mockup to test the method. RESULTS: The position of the pivot point is estimated with an average error less than 1.85 mm using two webcams placed from approximately 30 cm to 1 m away from the scene. The entire procedure was completed in a few seconds. CONCLUSION: In automated method to estimate the ideal position of the RCM was shown to be reliable. The method can be implemented within a visual servoing approach to automatically place the RCM point, or the results can be displayed on a screen to provide guidance to the surgeon. Further work includes the development of an image-guided alignment method and validation with in vivo experiments.

Does surface roughness influence the primary stability of acetabular cups? A numerical and experimental biomechanical evaluation.

Most acetabular cups implanted today are press-fit impacted cementless. Anchorage begins with the primary stability given by insertion of a slightly oversized cup. This primary stability is key to obtaining bone ingrowth and secondary stability. We tested the hypothesis that primary stability of the cup is related to surface roughness of the implant, using both an experimental and a numerical models to analyze how three levels of surface roughness (micro, macro and combined) affect the primary stability of the cup. We also investigated the effect of differences in diameter between the cup and its substrate, and of insertion force, on the cups' primary stability. The results of our study show that primary stability depends on the surface roughness of the cup. The presence of macro-roughness on the peripheral ring is found to decrease primary stability; there was excessive abrasion of the substrate, damaging it and leading to poor primary stability. Numerical modeling indicates that oversizing the cup compared to its substrate has an impact on primary stability, as has insertion force.

Understanding soft-tissue behavior for application to microlaparoscopic surface scan.

This paper presents an approach for understanding the soft-tissue behavior in surface contact with a probe scanning the tissue. The application domain is confocal microlaparoscopy, mostly used for imaging the outer surface of the organs in the abdominal cavity. The probe is swept over the tissue to collect sequential images to obtain a large field of view with mosaicking. The problem we address is that the tissue also moves with the probe due to its softness; therefore, the resulting mosaic is not in the same shape and dimension as traversed by the probe. Our approach is inspired by the finger slip studies and adapts the idea of load-slip phenomenon that explains the movement of the soft part of the finger when dragged on a hard surface. We propose the concept of loading-distance and perform measurements on beef liver and chicken breast tissues. We propose a protocol to determine the loading-distance prior to an automated scan and introduce an approach to compensate the tissue movement in raster scans. Our implementation and experiments show that we can have an image mosaic of the tissue surface in a desired rectangular shape with this approach.

Bone texture analysis is correlated with three-dimensional microarchitecture and mechanical properties of trabecular bone in osteoporotic femurs.

Fracture of the proximal femur is a major public health problem in elderly persons. It has recently been suggested that combining texture analysis and bone mineral density measurement improves the failure load prediction in human femurs. In this study, we aimed to compare bone texture analysis with three-dimensional (3D) microarchitecture and mechanical properties of trabecular bone in osteoporotic femurs. Eight femoral heads from osteoporotic patients who fractured their femoral neck provided 31 bone cores. Bone samples were studied using a new high-resolution digital X-ray device (BMA™, D3A Medical Systems) allowing for texture analysis with fractal parameter H (mean), and were examined using micro-computed tomography (microCT) for 3D microarchitecture. Finally, uniaxial compression tests to failure were performed to estimate failure load and apparent modulus of bone samples. The fractal parameter H (mean) was strongly correlated with bone volume fraction (BV/TV) (r = 0.84) and trabecular thickness (Tb.Th) (r = 0.91) (p < 0.01). H (mean) was also markedly correlated with failure load (r = 0.84) and apparent modulus (r = 0.71) of core samples (p < 0.01). Bone volume fraction (BV/TV) and trabecular thickness (Tb.Th) demonstrated significant correlations with failure load (r = 0.85 and 0.72, respectively) and apparent modulus (r = 0.72 and 0.64, respectively) (p < 0.01). Overall, the best predictors of failure load were H (mean), bone volume fraction, and trabecular thickness, with r (2) coefficients of 0.83, 0.76, and 0.80 respectively. This study shows that the fractal parameter H (mean) is correlated with 3D microCT parameters and mechanical properties of femoral head bone samples, which suggests that radiographic texture analysis is a suitable approach for trabecular bone microarchitecture assessment in osteoporotic femurs.

Building large mosaics of confocal edomicroscopic images using visual servoing.

Probe-based confocal laser endomicroscopy provides real-time microscopic images of tissues contacted by a small probe that can be inserted in vivo through a minimally invasive access. Mosaicking consists in sweeping the probe in contact with a tissue to be imaged while collecting the video stream, and process the images to assemble them in a large mosaic. While most of the literature in this field has focused on image processing, little attention has been paid so far to the way the probe motion can be controlled. This is a crucial issue since the precision of the probe trajectory control drastically influences the quality of the final mosaic. Robotically controlled motion has the potential of providing enough precision to perform mosaicking. In this paper, we emphasize the difficulties of implementing such an approach. First, probe-tissue contacts generate deformations that prevent from properly controlling the image trajectory. Second, in the context of minimally invasive procedures targeted by our research, robotic devices are likely to exhibit limited quality of the distal probe motion control at the microscopic scale. To cope with these problems visual servoing from real-time endomicroscopic images is proposed in this paper. It is implemented on two different devices (a high-accuracy industrial robot and a prototype minimally invasive device). Experiments on different kinds of environments (printed paper and ex vivo tissues) show that the quality of the visually servoed probe motion is sufficient to build mosaics with minimal distortion in spite of disturbances.