1208 Part F Field and Service Robotics pendent.Common sterilization methods include gamma Medical image segmentation and image fusion to rays (for disposable tools),autoclaving,soaking or gas construct and update patient-specific anatomic mod- sterilization,and the use of sterile drapes to cover un- els sterile components.Soaking or gas sterilization are less Biomechanical modeling for analyzing and pre- likely to damage robot components,but very rigorous dicting tissue deformations and functional factors cleaning is required to prevent extraneous foreign matter affecting surgical planning,control,and rehabilita- from shielding microbes from the sterilizing agent. tion Careful attention to higher levels of application pro- ● Optimization methods for treatment planning and tocols is also essential.Just like any other tool,surgical interactive control of systems robots must be used correctly by surgeons,and care- Methods for registering the virtual realiry of images ful training is essential for safe practice.Surgeons must and computational models to the physical realiry of understand both the capabilities and limitations of the an actual patient system.In surgical CAD/CAM applications,the sur- Methods for characterizing treatment plans and indi- Part geon must understand how the robot will execute the vidual task steps such as suturing,needle insertion, plan and be able to verify that the plan is being fol- or limb manipulation for purposes of planning,mon- 可 lowed.If the surgeon is interactively commanding the itoring,control,and intelligent assistance robot,it is essential that the robot interpret these com- Real-time data fusion for such purposes as updating mands correctly.Similarly,it is essential that the robot's models from intraoperative images model of its task environment correspond correctly to Methods for human-machine communication,in- the actual environment.Although careful design and im- cluding real-time visualization of data models, plementation can practically eliminate the likelihood of natural language understanding,gesture recognition, a runaway condition by the manipulator,this will do little etc. good if the robot is badly registered to the patient im- Methods for characterizing uncertainties in data, ages used to control the procedure.If the robot fails for models,and systems and for using this information any reason,there must be well-documented and planned in developing robust planning and control methods procedures for recovery (and possibly continuing the procedure manually). An in-depth examination of this research is beyond the Finally,it is important to remember that a well- scope of this article.A more complete discussion of these designed robot system can actually enhance patient topics may be found in the suggested further reading in safety.The robot is not subject to fatigue or momen- Sect.52.4. tary lapses of attention.Its motions can be more precise and there is less chance that a slip of the scalpel may 52.2.6 Registration damage some delicate structure.In fact,the system can be programmed to provide virtual fixtures(Sect.52.2.3) Geometric relationships are fundamental in medical preventing a tool from entering a forbidden region unless robotics,especially in surgical CAD/CAM.There is the surgeon explicitly overrides the system. an extensive literature on techniques for coregistering coordinate systems associated with robots,sensors,im- 52.2.5 Imaging and Modeling of Patients ages,and the patient [52.39,40].Following [52.40],we briefly summarize the main concepts here.Suppose that As the capabilities of medical robots continue to evolve, we have coordinates the use of computer systems to model dynamically changing patient-specific anatomy will become increas- A=(xA,yA,ZA) ingly important.There is a robust and diverse research 唱=(xB,yB,B), community addressing a very broad range of research topics.including the creation of patient-specific mod- corresponding to comparable locations in two coor- els from medical images,techniques for updating these dinate systems RefA and RefB.Then the process of models based upon real-time image and other sensor registration is simply that of finding a function TAB(...) data,and the use of these models for planning and mon- such that itoring of surgical procedures.Some of the pertinent research topics include the following: UB TAB(vA).1208 Part F Field and Service Robotics pendent. Common sterilization methods include gamma rays (for disposable tools), autoclaving, soaking or gas sterilization, and the use of sterile drapes to cover un￾sterile components. Soaking or gas sterilization are less likely to damage robot components, but very rigorous cleaning is required to prevent extraneous foreign matter from shielding microbes from the sterilizing agent. Careful attention to higher levels of application pro￾tocols is also essential. Just like any other tool, surgical robots must be used correctly by surgeons, and care￾ful training is essential for safe practice. Surgeons must understand both the capabilities and limitations of the system. In surgical CAD/CAM applications, the sur￾geon must understand how the robot will execute the plan and be able to verify that the plan is being fol￾lowed. If the surgeon is interactively commanding the robot, it is essential that the robot interpret these com￾mands correctly. Similarly, it is essential that the robot’s model of its task environment correspond correctly to the actual environment. Although careful design and im￾plementation can practically eliminate the likelihood of a runaway condition by the manipulator, this will do little good if the robot is badly registered to the patient im￾ages used to control the procedure. If the robot fails for any reason, there must be well-documented and planned procedures for recovery (and possibly continuing the procedure manually). Finally, it is important to remember that a well￾designed robot system can actually enhance patient safety. The robot is not subject to fatigue or momen￾tary lapses of attention. Its motions can be more precise and there is less chance that a slip of the scalpel may damage some delicate structure. In fact, the system can be programmed to provide virtual fixtures (Sect. 52.2.3) preventing a tool from entering a forbidden region unless the surgeon explicitly overrides the system. 52.2.5 Imaging and Modeling of Patients As the capabilities of medical robots continue to evolve, the use of computer systems to model dynamically changing patient-specific anatomy will become increas￾ingly important. There is a robust and diverse research community addressing a very broad range of research topics, including the creation of patient-specific mod￾els from medical images, techniques for updating these models based upon real-time image and other sensor data, and the use of these models for planning and mon￾itoring of surgical procedures. Some of the pertinent research topics include the following: • Medical image segmentation and image fusion to construct and update patient-specific anatomic mod￾els • Biomechanical modeling for analyzing and pre￾dicting tissue deformations and functional factors affecting surgical planning, control, and rehabilita￾tion • Optimization methods for treatment planning and interactive control of systems • Methods for registering the virtual reality of images and computational models to the physical reality of an actual patient • Methods for characterizing treatment plans and indi￾vidual task steps such as suturing, needle insertion, or limb manipulation for purposes of planning, mon￾itoring, control, and intelligent assistance • Real-time data fusion for such purposes as updating models from intraoperative images • Methods for human–machine communication, in￾cluding real-time visualization of data models, natural language understanding, gesture recognition, etc. • Methods for characterizing uncertainties in data, models, and systems and for using this information in developing robust planning and control methods An in-depth examination of this research is beyond the scope of this article. A more complete discussion of these topics may be found in the suggested further reading in Sect. 52.4. 52.2.6 Registration Geometric relationships are fundamental in medical robotics, especially in surgical CAD/CAM. There is an extensive literature on techniques for coregistering coordinate systems associated with robots, sensors, im￾ages, and the patient [52.39, 40]. Following [52.40], we briefly summarize the main concepts here. Suppose that we have coordinates vr A = (xA, yA,zA) vr B = (xB, yB,zB) , corresponding to comparable locations in two coor￾dinate systems RefA and RefB. Then the process of registration is simply that of finding a function TAB(···) such that vB = TAB(vA) . Part F 52.2