In this series of articles, we will look at several trends and directions in the application of robotic technology to health care. Featured in this review will be the use of robots in surgery, particularly surgery in extreme environments including the battlefield. Another area that we will look at is the current application of robots to non-surgical hospital care, including ICU and geriatric care. Another area will be acceptance issues, by patients and also by health care workers, of this new technology. We will also be delving into technical areas as well, and hopefully can discuss these areas in a way that is meaningful and educational to someone who wants to learn more about robots and health care.
Going beyond what is happening today, we will be looking at the future of health care robots. The future that I have in mind may be as little as ten years away, at least in some of its manifestations. However, there is no doubt that evolution of these future health care robots, which I call 'healing machines', will continue for a long time to come. This evolution is enabled by two factors, one technological and the other human. The technological factor is the availability of the basic machine abilities to enable these machines to function in a more or less human way. The core technology, of course, is the availability of very powerful, compact and relatively inexpensive computing engines to be the brain of these systems. Other technologies are equally important: among these are computer vision, voice recognition, and robotic manipulators with force feedback. All of these things have been around for some time, but they are just reaching the point where they are good enough, robust enough, and inexpensive enough, to be widely deployed. Another factor that is driving this evolution is the need of the military for automated combat casualty care. Robotics is an inherent part of the future of the military at least in the United States. Health care robotics is also going to be an inherent part of the future of manned space exploration, if we ever get around to doing that. For those reading this who are not fans of the military, let me say that it is a matter of history that advances in health care, particularly in surgical care, have resulted directly because of the need to care for wounded soldiers. This goes back at least to the time of Napoleon and continued notably in the twenty century (i.e. fluid resuscitation techniques and vascular surgery just to name two), and continues to this day. Regardless of anyone's ideology, the military is the organization that has not only the most acute need for improved surgical care, but also the money and the will to implement these advancements very quickly. The military's needs are truly life and death. Without a doubt, battlefield medicine and surgery have contributed to the improvement of health care for everyone, including soldiers and civilian patients ranging in age from neonatal to geriatric. We will talk about that story more in depth in a later article.
In my view, the healing machines of the future will be robots that are at least partly autonomous. This is in contrast to today's health care robots, which are basically machines that are teleoperated- directly controlled by the surgeon or doctor. The canonical example of such a teleoperated surgical robot is the well known and highly successful da Vinci from Intuitive Surgical. This is basically the only kind of surgical robot that we have today. The term 'teleoperated' is perhaps a little misleading, since the surgeon doing the 'tele'-operating is usually located quite close to the robot, in the same operating room just a few feet away. The real meaning of this word in this context is that the surgeon is directly controlling everything that the robot does. The robot is basically a sophisticated servo-mechanism that faithfully replicates the surgeon's hand and finger motions. The robot does not contribute anything original to the script, so to speak. It is literally an electromechanical puppet. The advantage of the robot comes from the fact that the surgeon's hand and finger motions can be translated into corresponding motions of the robot's manipulators, but these manipulators can be smaller, thinner, and generally less invasive than the surgeon's human hands and arms. This allows the surgery to be done on a minimally invasive basis, while still preserving at least some of the dexterity and flexibility of the best surgical tool ever invented- the human hand.
However, a more interesting and potentially much more useful kind of health care robot, that will be seen in the future, is one that can perform at least some of its functions without the need for direct, second by second, human input. We are not talking about performing these functions without human supervision but we do mean that the robot will be able to execute at least some of its therapeutic tasks without direct, motion by motion, step by step, human input, once certain parameters and goals have been defined. This type of control is called supervised autonomy. In case anyone is thinking that this is too farfetched, let me point out a well known, but seldom appreciated instance of supervised autonomy, which is in widespread clinical use today: that is LASIK surgery. In this type of surgery, a laser is used to reshape the surface of the cornea, to correct refractive errors and eliminate (or ameliorate) the need for corrective lenses. In this type of surgery, the actual firing and aiming of the laser is not under direct control of the human surgeon. It is doubtful that any but a few extremely skilled surgeons would have the necessary dexterity and precision of motion to apply the laser pulses in exactly the correct way to effect the desired changes in the surface of the cornea. Basically, only a machine can do that. Is this really surprising or shocking? The first transistor was assembled by hand in a laboratory by Shockley and his co-workers in the fifties. Today how many computer chips are built by hand? None. Forgetting about speed of assembly, extremely few if any humans would have the dexterity (or patience) to engrave the micro-tracings involved in a complex computer chip. But humans design the chip. In a similar way, a human surgeon designs the treatment plan for the cornea that needs reshaping, and reviews that plan prior to having the machine execute it. But once the machine starts to execute it, it is far better for the patient if the human surgeon steps back and lets the machine do its thing. Even today's LASIK machines, though not very smart, are a lot more precise than even the best human hand.
But other than delicate eye surgery and perhaps a few other examples, human surgeons are quite capable of performing the necessary tissue manipulations and obtaining good results. So why not stick with the teleoperational model embodied in the da Vinci robot? The teleoperated model can even be applied to surgery on remote battlefields- provided that the distance between the controlling surgeon and the controlled robot is not too great. And that is the catch. 'Latency' is how that issue here is referred to in surgical robotic circles. It basically means that as the distance increases, and becomes on the order of thousands of miles, the time that it takes for radio waves to travel from the controller to the controlled, becomes appreciable. That appreciable delay means that whatever the surgeon is doing and seeing actually happened some time ago. If this latency is only a few or tens of milliseconds, this is not an issue. But as this latency approaches a second or so, it does become an issue. The surgeon is able to compensate (humans are so good at adapting!) by operating more slowly. But there is a point at which this is not practical if the surgery is to be finished in a safe and expeditious matter. Latency is an issue even at terrestrial distances, but imagine what happens if the teleoperated robot is located on a space ship that is orbiting Mars. At certain times in Mars's orbit, the round trip time for a radio signal from earth (commanding the robot to do something ) to reach Mars and then return (bringing back video data about the effect that the action had) is about 45 minutes. This is far too long a delay for performing surgery. For instance, even in making a simple incision, a bleeding vessel may be encountered that needs to be tied off or cauterized. It would unacceptable to have to wait 45 minutes or so to take that action. The only way around that, other than having a human surgeon present, is to enable the robot to perform steps of the operation autonomously, using local feedback from its built in vision and tactile sensors. In this model, the human surgeon would take his/her time to look at the video and other data indicating what is wrong with the patient and where the incision needs to be made. Then the human teleoperator would instruct the robot where and exactly how to make the incision. The robot would review the plan with the surgeon and then get permission to execute it. This is pretty analogous to the LASIK machine being allowed to do the corneal surgery, once the plan has been approved by the human surgical supervisor. The difference between LASIK, however, (and this is a key difference) is that our future surgical robot would have enough sensor capability, smarts and dexterity to handle the 'routinely unexpected', for example a bleeding vessel that is in the path of the incision. This in fact is not too different in principle from the way an experienced attending surgeon will supervise a surgical resident in performing an operation. The resident, depending on level of skill, has enough sensor capability, smarts and dexterity to handle the 'routinely unexpected'. So the attending can just sit back and let the resident do the right thing—unless something really unexpected happens that requires the attending surgeon's experience and judgment. If something unexpectedly unexpected happens, a good resident will have a safe fall back position 'programmed' into his/her repertoire and will be able to hold the fort until the attending is available for input. The attending-resident model is an excellent model for combining machines and humans in surgery, since this model exploits the best features of the man and of the machine. Humans are excellent (or should be) at performing judgmental and evaluational tasks. Machines are excellent at performing technical maneuvers accurately and precisely.
Penelope is a small step toward the goal of supervised autonomy for robots in surgery. I feel this is true even though Penelope does not really do surgery, but merely dispenses and keeps track of surgical instruments. Unlike other robots in the operating room today, she really does have an independent 'brain', embodied in software as a rule-based inference engine. This independence is quite limited, of course, but she is the first robot in the operating room that includes such a capability. Another feature of Penelope which is important in the further development of autonomous surgical robots is the use of machine-vision. She is the first robot in the operating room to use this ability. Though limited to recognizing surgical instruments, this visual capability will someday extend to recognizing anatomical structures, a necessary ability if the robot is to be able to perform surgery in a partly autonomous way.
In summary, this upcoming series of articles will examine and comment on many interesting issues, some social and some technological, that have to do with the use of robots in health care, medical and surgical...from present day teleoperated robots to the true healing machines of the future.