Moving Past Normal Force: Capturing and Classifying Shear Motion Using 3D Sensors

Abstract

In our previous research, we used clinical breast examination models instrumented with direct (normal) force sensors for training and assessment. A weakness of the normal force sensors is the ability to delineate, in detail, all of the performance measures we wish to understand. This study incorporated the use of newly developed shear force sensors to extend a framework for quantifying hands-on performance.

1. Introduction

A clinical breast examination can be defined by a set of hand movements known as exploratory procedures (EPs). As described by Klatzky and Lederman, EPs are a set of reproducible and subconscious hand motions used by humans to explore object properties [1, 2]. For example, the pressure EP, which is defined as applying a normal force to an object, is used to explore hardness. The lateral motion EP is used to explore an object that has texture. In our previous work, we have shown that clinical breast examination models, instrumented with force sensing resistors (FSR’s), can be used to quantify hands-on performance [3]. In addition, there is evidence that EPs can be detected during simulated breast examinations [4]. However, FSR sensors cannot directly delineate between the various types of EPs. FSR sensors only measure normal force, which is useful in capturing the pressure EP. However, FSR sensors do not capture the shear forces induced by the lateral motion EP, a major component of the clinical breast exam. The goal of this study was to use shear force sensors to extend a framework for delineating between pressure and lateral motion EPs. Specifically, this study investigated two types of shear force sensors: a “3D sensor” and an “Artificial Haircell sensor.” The study aims to show how both sensors can be used in conjunction with an FSR sensor to provide a more complete mechanism for quantifying clinical breast exam performance.

2. Methods and Materials

2.1. Sensor Setup

The Artificial Haircell sensor can measure shear force with a single microelectronic cantilever, Figure 1 (left). Applying different shear forces on the cantilever produces different electronic signals on a single output. The 3D sensor is a microelectronic sensor with two cantilevers, one for horizontal shear and one for vertical shear, Figure 1 (right). As such, the 3D sensor electronically reports shear forces on two separate outputs.

Figure 1.

Close-up of Artificial Haircell sensor (left) and the 3D sensor with axes (right)

Tests performed on the 3D sensor matched the conditions used for the Haircell, including investigating shear forces and placing a .2″ radius FSR over the 3D sensor.

2.2. Data Analysis

The 3D sensor was adjusted to output channel voltages before recording. The first 3 seconds of recording were reserved for baseline voltage calculations, and shear forces were applied to the sensors during the next 10 seconds. The first 30 samples of each recording were averaged and subtracted from its corresponding 3D sensor column, to form baseline voltages. The baseline voltages were multiplied by 10 to enhance data clarity.

2.3. Sensor Characterization

Tests comprised of recording rubbing motions in the horizontal (x-axis) and vertical (y-axis) directions to induce horizontal and vertical shear forces on a Polydimethylsiloxane (PDMS) surface of the 3D sensor. Motions that move along the direction of the wires were defined as vertical and motions across the wires were defined as horizontal.

Tests were conducted on the center of the 3D sensor. In the recording of vertical and horizontal rubbing, hand motions were used to capture unilateral motions. To reduce interference between vertical and horizontal motions, movements were kept on one side of the adjacent plane. For example, horizontal motions were conducted on the vertical plane’s positive axis. The expected plot would show the horizontal plane with positive and negative readings (indicating motions in the positive and negative x directions) while keeping the vertical readings positive (indicating motions only on one side of the sensor).

2.4. Verification

Shear motion capture in our breast models would be verified through analyzing sensor data from a participant’s exam, recorded during the 2011 American Society of Clinical Oncology (ASCO) conference. This recording was conducted with a prosthetic breast, breast skin, and PDMS pathology over the 3D sensor. Analysis of the exam waveform and the associated video recording was conducted.

Plots for each data set displayed normal and shear force recordings. For the ASCO participant plot, Figure 4, the black line corresponds with motions corresponding to the x-axis (horizontal readings), and gray line for y-axis (vertical readings). The light gray line corresponds to recordings by the FSR. For the Haircell, the black line corresponds with Haircell readings, and the light gray with FSR readings.

Figure 4.

Waveform from the clinician using the clinical breast exam simulator

3. Results

3.1. Sensor Characterization

Both sensors showed that the shear forces produced by lateral motions could be captured. The 3D sensor recordings had shown distinct differences, namely left and right motions displayed only positive or only negative readings. In contrast, the Haircell display produced similar waveforms for left and right motions.

These distinct differences can be seen in vertical and horizontal shear force plots. For horizontal rubbing on the 3D sensor, it is observed that x direction readings cross the x-axis, while the y readings stay positive. This confirmed that the hand motions involved were horizontal movements across the upper half of the 3D sensor (Y’s positive section) (Figure 2).

Figure 2.

Haircell (Left) and 3D Sensor (Right) comparisons. Top plots refer to vertical rubbing (y-axis) and bottom plots refer to horizontal rubbing (x-axis).

For vertical rubbing, it is observed that the 3D sensor’s y readings cross the x-axis, supporting the sensor’s ability to record vertical shear force. This plot was identified as vertical rubbing by its greater voltage variations in the vertical direction rather than the horizontal readings. This supports the view that vertical recordings were made, and that horizontal recordings were an incidental result (Figure 2). In contrast, it is difficult to determine the difference between vertical and horizontal plots of the Haircell. Due to this difficulty and cost considerations, the 3D sensor was selected over the Haircell sensor for use in future studies.

3.2. Verification

Verification of shear force capture can be observed in Figures 3 and 4. Shear forces induced by the participant’s hand motions can be identified, with waveforms matching the hand motions seen in the video, Figure 3. The incorporation of the 3D sensor allows for the characterization of shear motions of the hand.

Figure 3.

Screenshot of a clinician performing the clinical breast exam on our simulator instrumented with the 3D sensor.

4. Discussion

This study demonstrated that 3D sensors could detect the shear forces caused by rubbing. Using 3D sensors, we extend the Klatzky and Lederman characterization of hand motions to include the ability to measure shear force. Measuring shear forces through the use of 3D sensors could contribute to our understanding of hand motions during a clinical breast exam.

Furthermore, incorporation of 3D sensors into breast models can further aid in future efforts to standardize clinical breast examination performance, an improvement that could not be done through the use of only FSRs. The 3D sensor can be used in conjunction with an FSR sensor to help extend the framework for delineating between EP’s through the direct quantification of shear and normal forces, thereby providing a more complete performance feedback mechanism for quantifying the clinical breast exam.

In addition, we decided to replace the Artificial Haircell sensor with the 3D sensor in our Clinical Breast Examination Models, due to signal analysis clarity and cost considerations [5, 6].

As future work, more studies will be required to better understand the use of the 3D sensor in the breast model. The effects of potentially confounding variables such as breast prosthetic and skin interaction with the 3D sensor are not known.