Policy search methods can allow robots to learn control policies for a wide range of tasks, but practical applications of policy search often require hand-engineered components for perception, state estimation, and low-level control. In this paper, we aim to answer the following question: does training the perception and control systems jointly end-to-end provide better performance than training each component separately? To this end, we develop a method that can be used to learn policies that map raw image observations directly to torques at the robot's motors. The policies are represented by deep convolutional neural networks (CNNs) with 92,000 parameters, and are trained using a guided policy search method, which transforms policy search into supervised learning, with supervision provided by a simple trajectory-centric reinforcement learning method. We evaluate our method on a range of real-world manipulation tasks that require close coordination between vision and control, such as screwing a cap onto a bottle, and present simulated comparisons to a range of prior policy search methods.

/pdf/end-to-end-training-of-deep-visuomotor-policies-1vqd5uo50d.pdf

End-to-end training of deep visuomotor policies

Robotic surgery is a new and exciting emerging technology that is taking the surgical profession by storm. Up to this point, however, the race to acquire and incorporate this emerging technology has primarily been driven by the market. In addition, surgical robots have become the entry fee for centers wanting to be known for excellence in minimally invasive surgery despite the current lack of practical applications. Therefore, robotic devices seem to have more of a marketing role than a practical role. Whether or not robotic devices will grow into a more practical role remains to be seen.

Our goal in writing this review is to provide an objective evaluation of this technology and to touch on some of the subjects that manufacturers of robots do not readily disclose. In this article we discuss the development and evolution of robotic surgery, review current robotic systems, review the current data, discuss the current role of robotics in surgery, and finally we discuss the possible roles of robotic surgery in the future. It is our hope that by the end of this article the reader will be able to make a more informed decision about robotic surgery before “chasing the market.”

Robotic Surgery: A Current Perspective

Micro- and nanoscale robots that can effectively convert diverse energy sources into movement and force represent a
rapidly emerging and fascinating robotics research area. Recent advances in the design, fabrication, and operation of micro/nanorobots have greatly enhanced their power, function, and versatility. The new capabilities of these tiny untethered machines indicate immense potential for a variety of biomedical applications. This article reviews recent progress and future perspectives of micro/nanorobots in biomedicine, with a special focus on their potential advantages and applications for directed drug delivery, precision surgery, medical diagnosis, and detoxification. Future success of this technology, to be realized through close collaboration between robotics, medical, and nanotechnology experts, should have a major impact on disease diagnosis, treatment, and prevention.

/pdf/micro-nanorobots-for-biomedicine-delivery-surgery-sensing-23eglg29gr.pdf

Micro/Nanorobots for Biomedicine: Delivery, Surgery, Sensing, and Detoxification.

Any device which senses information such as shape, texture, softness, temperature, vibration or shear and normal forces, by physical contact or touch, can be termed a tactile sensor. The importance of tactile sensor technology was recognized in the 1980s, along with a realization of the importance of computers and robotics. Despite this awareness, tactile sensors failed to be strongly adopted in industrial or consumer markets. In this paper, previous expectations of tactile sensors have been reviewed and the reasons for their failure to meet these expectations are discussed. The evolution of different tactile transduction principles, state of art designs and fabrication methods, and their pros and cons, are analyzed. From current development trends, new application areas for tactile sensors have been proposed. Literature from the last few decades has been revisited, and areas which are not appropriate for the use of tactile sensors have been identified. Similarly, the challenges that this technology needs to overcome in order to find its place in the market have been highlighted.

/pdf/a-review-of-tactile-sensing-technologies-with-applications-12ul5smpoy.pdf

A review of tactile sensing technologies with applications in biomedical engineering

Policy search methods can allow robots to learn control policies for a wide range of tasks, but practical applications of policy search often require hand-engineered components for perception, state estimation, and low-level control. In this paper, we aim to answer the following question: does training the perception and control systems jointly end-to-end provide better performance than training each component separately? To this end, we develop a method that can be used to learn policies that map raw image observations directly to torques at the robot's motors. The policies are represented by deep convolutional neural networks (CNNs) with 92,000 parameters, and are trained using a partially observed guided policy search method, which transforms policy search into supervised learning, with supervision provided by a simple trajectory-centric reinforcement learning method. We evaluate our method on a range of real-world manipulation tasks that require close coordination between vision and control, such as screwing a cap onto a bottle, and present simulated comparisons to a range of prior policy search methods.

End-to-End Training of Deep Visuomotor Policies

Minimally invasive surgical (MIS) procedures using long instruments have profoundly influenced modern surgery by decreasing invasiveness, therefore minimizing patient recovery time and cost. However, surgical procedures using long tools inserted through small ports on the body deprive surgeons of the depth perception, dexterity, sense of touch, and straightforward hand-eye coordination that they are accustomed to in open procedures. Surgeons have traditionally relied on the combination of their tactile and visual abilities to diagnose tissue as normal or abnormal.1

The use of laparoscopic tools capable of providing force feedback is lacking in current robot-assisted surgical systems. Current robotic surgical systems, while providing excellent visual feedback, are incapable of providing force feedback. Surgeons who operate using these robotic assistants do not have force and tactile feedback from the operative site as they are now indirectly in contact with the surgical site instrumented tools. Therefore, the need to incorporate force feedback capabilities into MIS procedures, especially robotically assisted procedures, provides an excellent opportunity to improve the quality of surgical procedures. Improvements in MIS systems will lead to significant societal impacts through better patient care, reduced morbidity, shorter hospital stays, reduced trauma, faster recovery times, and lower health care costs.

While research has been done on developing laparoscopic tools capable of providing force feedback, their use in surgical practice is limited by the size of these instrumented laparoscopic tools and the ease of use in surgical procedures. As a result of these shortcomings, our goal was to (a) design and develop a modular laparoscopic tool, which would provide force feedback for tissue characterization; and (b) using the developed laparoscopic grasper, we wanted to evaluate the role of force feedback in MIS. In this paper, we will focus primarily on the second goal, namely, the evaluation of the role of force feedback in MIS. We have developed an interface to aid the surgeons in characterizing artificial tissues of varying stiffness (prepared using hydrogels), which are a reasonable approximation to the tissues encountered by surgeons in liver tumor resection. The soft tissues in our experiments will approximately represent normal liver tissue, medium tissues will approximately represent a tumor in the formation stage, and hard tissues will approximately represent a fully developed tumor. These tissue samples were tested by palpation by our surgeon collaborator to make a realistic choice of artificial tissue samples in our experimental studies.

Several researchers have already proposed solutions to incorporate force feedback into current laparoscopic tools.2–8 Many of these solutions involve adding force sensors, such as strain gauges, to record forces at the tool tip or developing a new robotic manipulator used to operate laparoscopic tools and obtain force feedback. Additional work has been done in creating new laparoscopic tools with force feedback incorporated into their designs.9–12 However, there still exist many problems within the designs of laparoscopic tools and their use in robotic surgery.13,14 Schurr et al15 have developed a 2-arm, master-slave manipulator called ARTEMIS for use by surgeons. Dario et al16 have developed a miniature robotic surgical system for colonoscopy. Several master-slave robotic surgical systems have also been commercially developed. However, these systems do not provide force feedback to the surgeon. Current research within the area of force feedback has led to the use of cable-driven mechanisms to reduce backlash and play in the mechanism.17–19

Force feedback plays a significant role in minimally invasive surgery: results and analysis.

One of the limitations of current surgical robots used in surgery is the lack of haptic feedback. While current surgical robots improve surgeon dexterity, decrease tremor, and improve visualization, they lack the necessary fidelity to help a surgeon characterize tissue properties for improving diagnostic capabilities. Our work focuses on the development of tools and software that will allow haptic feedback to be integrated in a robot-assisted gastrointestinal surgical procedure. In this paper, we have developed several tissue samples in our laboratory with varying hardness to replicate real-tissues palpated by a surgeon in gastrointestinal procedures. Using this tissue, we have developed a novel setup whereby the tactile feedback from the laparoscopic tool is displayed on the PHANToM haptic interface device in real-time. This is used for tissue characterization and classification. Several experiments were performed with different users and they were asked to identify the tissues. The results demonstrate the feasibility of our approach.

/pdf/real-time-haptic-feedback-in-laparoscopic-tools-for-use-in-4v5ovhxid3.pdf

Real-Time Haptic Feedback in Laparoscopic Tools for Use in Gastro-Intestinal Surgery

Formerly, most of the causes and treatments of chronic lower abdominal and groin pain in high-performance athletes eluded sports medicine specialists. Now we are much better at identifying and managing the different syndromes. Most of the advances are based on empiric evidence, although many pitfalls remain with respect to diagnosis and management of the various syndromes. This article reviews our current understanding of several of these clinical entities. We focus on lessons learned from a large 15-year experience.

Surgical Management of Chronic Lower Abdominal and Groin Pain in High-performance Athletes

Background
One of the most concerning and potentially fatal complications of gastric bypass surgery is pulmonary embolism (PE) with published rates exceeding 1%. Although this procedure has been proven effective in reducing the morbidity and mortality of obesity and associated health care costs, it is elective and risk must be minimized. No dosing guidelines exist for pharmacologic prophylaxis in obese patients who are already at increased risk for these events. Although the current ASMBS and Chest guidelines recommend preoperative pharmacologic prophylaxis against thromboembolic events, no standard dosing protocols exist for the obese population. We propose a protocol including immediate pre followed by twice daily postoperative BMI-based dosing of low-molecular-weight heparin (LMWH), along pneumatic compression devices, and early ambulation.

Evaluating the Safety and Efficacy of BMI-Based Preoperative Administration of Low-Molecular-Weight Heparin in Morbidly Obese Patients Undergoing Roux-en-Y Gastric Bypass Surgery

Andres Castellanos

Papers

Robotic Surgery: A Current Perspective

Force feedback plays a significant role in minimally invasive surgery: results and analysis.

Real-Time Haptic Feedback in Laparoscopic Tools for Use in Gastro-Intestinal Surgery

Surgical Management of Chronic Lower Abdominal and Groin Pain in High-performance Athletes

Evaluating the Safety and Efficacy of BMI-Based Preoperative Administration of Low-Molecular-Weight Heparin in Morbidly Obese Patients Undergoing Roux-en-Y Gastric Bypass Surgery