**2.1 Force sensorized cutting tool**

Typical cutting forces in surgical interventions can reach values of up to ±3N for tangential and 10N for normal loadings. The device described in (Valdastri et al., 2005) uses a siliconbased MEMS sensor to measure forces on a cutting blade. The MEMS sensor deployed in the

Sensorized Tools for Haptic Force Feedback in Computer Assisted Surgery 137

After these parts were cut out and prepared the cutting blade is brazed to the stainless steel structure using a special holding fixture. The cutting blade used was a conventional scalpel blade cut with wire EDM to smaller dimensions, both to reduce the size of the overall tool and also to allow a proper fit to the other parts of the sensor. To measure the deflection of the 3 support arms, micro fabricated strain gauges from Vishay on a flexible substrate were then bonded to the structure with M-600 epoxy from Vishay. For assembly of the first prototype separate micro strain gauges were used to simplify the construction. These strain gauges were of an overall area of approximately 2mm x 2mm (contact pads included, foil grid of 380µm x 500 µm) and excess material was cut out to leave a strain gauge of approximately 450µm x 1000µm. In reality mass production requires an automatized industrial assembly process but for the first prototype the manual steps are sufficient. In the final design the three micro strain gauges are incorporated with the necessary connections on a common flexible substrate, which would make assembly much easier (figure 1b).

Once the strain gauges were mounted (Fig 2a) , the next step was to bond wires to the strain gauges by soldering and then set the wires so that they can be easily taken out of the tool for connection to a suitable connector. To house the sensor an aluminum tube of diameter 4mm was chosen. Recesses were machined into the top part of the tube to allow mounting of the sensor tip. The sensor tip is then carefully placed into the tube with all wires and then set into a fixture. The final step is to form the top lip of the aluminum tube over the sensor, thus partially closing the sensor and mechanically clamping it to the tube. For the prototype a manual fixture was used to close the structure but for a more automated approach with higher tolerances and overall quality a customized press would be necessary. A Polyoxymethylene (POM) support was developed for a first characterisation of the device

a b

Figure 3 details the processing of the strain gauges (Rsg1, Rsg2 and Rsg3) in ¼ Wheatstone bridge configuration. A reference voltage of 2.5 V is generated with the two resistors Rref1

Fig. 2. a. Micro strain gauges mounted on the steel structure; b. Demonstrator ready for

and Rref 2 with 10 kΩ each. Three potentiometers are integrated for offset removal.

(Fig 2b).

characterisation

final prototype was designed for about one tenth of these forces, thus a kinematical chain was introduced to scale down the forces on the cutting blade to the MEMS force range.

Unfortunately the viscoelastic behavior of the polyurethane kinematical chain allowed repeatable results only for alternating loadings, but not for static ones. Moreover the limited force resolution of 50mN was not sufficient for a precise haptic force feedback. Additionally the tool turned out to be quite fragile: Applying strong tangential forces to the blade often resulted in a detachment of the blade support and breakage of the silicon structure. While further developments of that prototype are still ongoing, in parallel to that, a complete redesign of the force sensorized cutting tool was initiated with focus on:


The main idea behind the development of the novel tool is to design a mechanical structure from a strong and tough material (stainless steel) that deflects linearly when a normal or tangential force is applied in the required force range. On the support micro strain gauges are bonded, which allows the direct measurement of the elastic deflections of the structure thus simplifying the whole device. Thus a kinematical chain as in (Valdastri et al., 2005) can be avoided (the viscoelastic behavior of the kinematical chain was the main reason for inaccurate readings for static forces).

Figure 1a details the first design of the cutting tool. Mechanical simulations were carried out in Comsol Multyphysics (figure 1a) to calculate optimum dimensions, whilst keeping the overall size to a minimum and mechanical performance to a maximum. The most important criteria are that the sensor can function to a maximum load of ±3N tangential and ±20 N normal and that the force readings are linear and have low hysteresis. The first structure prototype parts were fabricated using micro-Wire-EDM from a plate of 400µm thick stainless steel.

Fig. 1. a. FEMLAB simulation; b. Tool tip design

final prototype was designed for about one tenth of these forces, thus a kinematical chain was introduced to scale down the forces on the cutting blade to the MEMS force range.

Unfortunately the viscoelastic behavior of the polyurethane kinematical chain allowed repeatable results only for alternating loadings, but not for static ones. Moreover the limited force resolution of 50mN was not sufficient for a precise haptic force feedback. Additionally the tool turned out to be quite fragile: Applying strong tangential forces to the blade often resulted in a detachment of the blade support and breakage of the silicon structure. While further developments of that prototype are still ongoing, in parallel to that, a complete

The main idea behind the development of the novel tool is to design a mechanical structure from a strong and tough material (stainless steel) that deflects linearly when a normal or tangential force is applied in the required force range. On the support micro strain gauges are bonded, which allows the direct measurement of the elastic deflections of the structure thus simplifying the whole device. Thus a kinematical chain as in (Valdastri et al., 2005) can be avoided (the viscoelastic behavior of the kinematical chain was the main reason for

Figure 1a details the first design of the cutting tool. Mechanical simulations were carried out in Comsol Multyphysics (figure 1a) to calculate optimum dimensions, whilst keeping the overall size to a minimum and mechanical performance to a maximum. The most important criteria are that the sensor can function to a maximum load of ±3N tangential and ±20 N normal and that the force readings are linear and have low hysteresis. The first structure prototype parts were fabricated using micro-Wire-EDM from a plate of 400µm thick

a b

redesign of the force sensorized cutting tool was initiated with focus on:



inaccurate readings for static forces).

Fig. 1. a. FEMLAB simulation; b. Tool tip design

stainless steel.


After these parts were cut out and prepared the cutting blade is brazed to the stainless steel structure using a special holding fixture. The cutting blade used was a conventional scalpel blade cut with wire EDM to smaller dimensions, both to reduce the size of the overall tool and also to allow a proper fit to the other parts of the sensor. To measure the deflection of the 3 support arms, micro fabricated strain gauges from Vishay on a flexible substrate were then bonded to the structure with M-600 epoxy from Vishay. For assembly of the first prototype separate micro strain gauges were used to simplify the construction. These strain gauges were of an overall area of approximately 2mm x 2mm (contact pads included, foil grid of 380µm x 500 µm) and excess material was cut out to leave a strain gauge of approximately 450µm x 1000µm. In reality mass production requires an automatized industrial assembly process but for the first prototype the manual steps are sufficient. In the final design the three micro strain gauges are incorporated with the necessary connections on a common flexible substrate, which would make assembly much easier (figure 1b).

Once the strain gauges were mounted (Fig 2a) , the next step was to bond wires to the strain gauges by soldering and then set the wires so that they can be easily taken out of the tool for connection to a suitable connector. To house the sensor an aluminum tube of diameter 4mm was chosen. Recesses were machined into the top part of the tube to allow mounting of the sensor tip. The sensor tip is then carefully placed into the tube with all wires and then set into a fixture. The final step is to form the top lip of the aluminum tube over the sensor, thus partially closing the sensor and mechanically clamping it to the tube. For the prototype a manual fixture was used to close the structure but for a more automated approach with higher tolerances and overall quality a customized press would be necessary. A Polyoxymethylene (POM) support was developed for a first characterisation of the device (Fig 2b).

Fig. 2. a. Micro strain gauges mounted on the steel structure; b. Demonstrator ready for characterisation

Figure 3 details the processing of the strain gauges (Rsg1, Rsg2 and Rsg3) in ¼ Wheatstone bridge configuration. A reference voltage of 2.5 V is generated with the two resistors Rref1 and Rref 2 with 10 kΩ each. Three potentiometers are integrated for offset removal.

Sensorized Tools for Haptic Force Feedback in Computer Assisted Surgery 139

microcontroller [Atmel, ATmega32] and then connected to the PC via a serial to USB interface [FT232, Future Technology Devices International Limited, United Kingdom]. The firmware for the microcontroller was developed in AVR Studio [Atmel] and the WinAVR GNU C Compiler. Measurement data are then processed and visualized with a GUI under LabView. Several prototypes with various lengths (figure 5a, b and 6 a, b) were manufactured and

a b

Fig. 5. a. Prototypes with an overall length of 25 and 50mm including the support structure;

a b

In a first characterisation of the device the output signals were compared to the ones of a commercial available load cell in a setup similar to (Valdastri et al., 2007). The sensor gives linear readings of strain gauge output versus load. The sensor is also robust which is necessary for surgical applications. Its small size (4mm diameter) makes it suitable for

characterized. In the following the main technical data of the prototypes are listed:

Length: 16mm (including support): Diameter of the device: 4mm

b. Prototype of total length 20 mm

robotic aided surgery.

Fig. 6. Final prototype has an outer diameter of 4 mm

Tangential forces: max ±3N, 4 mN resolution Normal forces: max 20 N, 50 mN resolution

For processing of the three analog channels a 24bit sigma delta AD converters [AD7730, Analog Devices] are used, which are with their programmable input range from 10mV to +- 80 mV especially designed for sensor signal processing. Reference voltages as low as 1V are possible. Moreover a programmable digital filter can be used to suppress noise.

Fig. 3. Strain gauges in ¼ Wheatstone Bridge configuration

Fig. 4. Interfacing the AD converters

In a first prototype of the electronics the AD converters were directly connected to the parallel port of a PC and then interfaced with software developed in LabView 8.5 [National Instruments]. The software emulated serial peripheral interface did not perform flawlessly, thus in the final version (Fig 4) the three analog AD converters are interfaced to a 8 bit RISC microcontroller [Atmel, ATmega32] and then connected to the PC via a serial to USB interface [FT232, Future Technology Devices International Limited, United Kingdom]. The firmware for the microcontroller was developed in AVR Studio [Atmel] and the WinAVR GNU C Compiler. Measurement data are then processed and visualized with a GUI under LabView.

Several prototypes with various lengths (figure 5a, b and 6 a, b) were manufactured and characterized. In the following the main technical data of the prototypes are listed:

Length: 16mm (including support): Diameter of the device: 4mm Tangential forces: max ±3N, 4 mN resolution Normal forces: max 20 N, 50 mN resolution

138 Haptics Rendering and Applications

For processing of the three analog channels a 24bit sigma delta AD converters [AD7730, Analog Devices] are used, which are with their programmable input range from 10mV to +- 80 mV especially designed for sensor signal processing. Reference voltages as low as 1V are

In a first prototype of the electronics the AD converters were directly connected to the parallel port of a PC and then interfaced with software developed in LabView 8.5 [National Instruments]. The software emulated serial peripheral interface did not perform flawlessly, thus in the final version (Fig 4) the three analog AD converters are interfaced to a 8 bit RISC

possible. Moreover a programmable digital filter can be used to suppress noise.

Fig. 3. Strain gauges in ¼ Wheatstone Bridge configuration

Fig. 4. Interfacing the AD converters

Fig. 5. a. Prototypes with an overall length of 25 and 50mm including the support structure; b. Prototype of total length 20 mm

Fig. 6. Final prototype has an outer diameter of 4 mm

In a first characterisation of the device the output signals were compared to the ones of a commercial available load cell in a setup similar to (Valdastri et al., 2007). The sensor gives linear readings of strain gauge output versus load. The sensor is also robust which is necessary for surgical applications. Its small size (4mm diameter) makes it suitable for robotic aided surgery.

Sensorized Tools for Haptic Force Feedback in Computer Assisted Surgery 141

early development state are very fragile and can be damaged easily, thus haptic force feedback is a important safety feature in order to avoid to damage to the vessels or the heart

The catheter prototype is fabricated from polyurethane with a flexible end of 40 mm length and an outer diameter of 3,5 mm. The diameter of the main inner lumen is 2,5 mm . Four 380 µm thin lumen are integrated in the wall (thickness: 550 µm) and serve as guidance for four Bowden cables. A micromachined tip (Fig 7a and 7b) fabricated from Polyether Ether Ketone (PEEK) is bonded to the flexible end of the catheter and mainly acts as a support for four impedance measurement electrodes and two optical fibers. Furthermore it offers a main channel leading to the inner lumen of the catheter. This inner lumen serves as a supply of

when navigation the catheter, which would be life threatening for the fetus.

a b

To be able to reach the point of interest, the catheter needs to be equipped with steering capabilities. The multi lumen catheter consists of a very flexible ending and a less flexible part. In the walls of the catheter four lumen are integrated, each one for one steering wire. Such wires are usually referred as Bowden cables. Pulling on these four wires and releasing at the same time the wire which is on the opposite side in the catheter will primarily result in a bending movement of the flexible end part of the catheter. Two microcontroller driven servo drives are used to pull and release the wires. This microcontroller is then connected to a personal computer, which is equipped with a haptic force feedback joystick allowing a precise control of the catheter. A third degree of freedom is realized by either manually or servo supported driving the catheter forward and backwards. The arm on the servo controller, which is actuating the Bowden wire is equipped with a strain gauge, which allows measurement of the actuation forces on the Bowden wire. As the actuation force reflects the force required to bend the catheter tip, it also reflects a force on the catheter tip itself, thus it can be used for haptic force feedback. For the readout of the strain gauges a similar circuit, but just with two AD converters, like described in 2.1, figure 4 was used.

Figure 8 shows the first prototype of the catheter controlled by a commercial joystick.

Fig. 7. a. Catheter tip prototype; b. Catheter tip 3D model

**2.2.3 Catheter design** 

washing solution (physiological saline solution).
