Towards Robotised Palpation for Cancer Detection through Online Tissue Viscoelastic Characterisation with a Collaborative Robotic Arm

TL;DR

The paper addresses online, real-time estimation of end-effector penetration $d$ and viscoelastic tissue parameters during robotic palpation, without relying on force sensors. It combines a dimensionality-reduced Hunt-Crossley contact model with a dynamic arm-tissue system and an Extended Kalman Filter to jointly estimate penetration and tissue parameters in two configurations: with force sensing and sensorless using impedance-based force approximation. Experiments on silicone phantoms with varying stiffness show that DRM-based Kalman filtering yields accurate estimates of stiffness $k_M$ or $\kappa$ and damping $c_M$ or $\lambda$, and can distinguish hard intrusions simulating cancer, with fast convergence. The approach promises aRobotic palpation framework capable of providing objective diagnostic cues for early cancer detection, potentially enhanced by ultrasound for depth localization and robustness.

Abstract

This paper introduces a new method for estimating the penetration of the end effector and the parameters of a soft body using a collaborative robotic arm. This is possible using the dimensionality reduction method that simplifies the Hunt-Crossley model. The parameters can be found without a force sensor thanks to the information of the robotic arm controller. To achieve an online estimation, an extended Kalman filter is employed, which embeds the contact dynamic model. The algorithm is tested with various types of silicone, including samples with hard intrusions to simulate cancerous cells within a soft tissue. The results indicate that this technique can accurately determine the parameters and estimate the penetration of the end effector into the soft body. These promising preliminary results demonstrate the potential for robots to serve as an effective tool for early-stage cancer diagnostics.

Figures (5)

• Figure 1: Schematic of a spherical indenter in constant with a viscoelastic half-plane. $\Delta x$ is the distance between the viscous elements, $a$ represents half of the projection of the circle's portion in contact, $d$ is the maximum penetration and $R_1$ is the equivalent radius of the sphere in the viscoelastic halfplane.
• Figure 2: The left model represents the system described in \ref{['eq:state_space_lin_FT']}, where it is assumed that a force is acting on the F/T sensor. The sensor is connected to a mass, which is in turn connected to a spring-damper element. The model described in \ref{['eq:im_sys']} is shown on the right, where the impedance control model is displayed instead of the external force.
• Figure 3: (a) The experimental setup is composed of a position-controlled Ur3e, a 6-axis force torque sensor BOTA SensorONe, a 3D printed indenter and a silicone specimen. (b) The 4 types of silicone. S1 is softer (ECOFLEX-0030), S2 is the soft silicone with the steel ball, S3 refers to the stiffer silicone (Dragonskin-10kN) and S4 is the same silicone as S3 with the steel ball.
• Figure 4: Results of the states estimation using the developed models with the softer silicone S1. In red with the label R are plotted the reference values computed with the LS method. In the first row are shown the results using the Kelvin-Voigt model with and without force sensor (M$_1$ and M$_2$). The second row shows the results of the estimation using the DM to model the soft body with and without the force torque sensor (M$_3$ and M$_4$).
• Figure 5: Comparison of the registered forces (F/T in red) against the estimated ones (M$_2$ in green and M$_4$) without the use of the sensor.