A Novel Soft and Inflatable Strain-based Tactile Sensing Balloon for Enhanced Diagnosis of Colorectal Cancer Polyps Via Colonoscopy

Abstract

In this paper, with the goal of addressing the lack of tactile feedback in colorectal cancer (CRC) polyps diagnosis using a colonoscopy procedure, we propose the design and fabrication of a novel soft and inflatable strain-based tactile sensing balloon (SI-STSB). The proposed soft sensor features a unique stretchable sensing layer – that utilizes a liquid metal injected within spiral-shape microchannels of a stretchable substrate – and is integrated with a unique inflatable balloon mechanism. The proposed SI-STSB has been thoroughly characterized through different calibration experiments. Results demonstrate a phenomenal adjustable sensitivity with low hysteresis behavior under different experimental conditions for this sensor making it a great candidate for enhancing the existing diagnosis procedures.

Graphical Abstract

I. Introduction

Occurrence of colorectal cancer (CRC) has significantly increased worldwide in recent years, making it the second leading cause of cancer-related death [1]. According to estimates based on population growth and aging, the number of new cases of CRC is projected to reach 3.2 million by 2040 [2]. However, if colorectal cancer is detected early, the survival rate would increase to 91% [3]. Therefore, it is crucial to promptly detect pre-cancerous lesions (polyps) using sensitive diagnostic techniques to decrease the risk of mortality and provide more treatment options. Recent studies have highlighted a significant correlation between a change in the elasticity of polyps and cancer cell invasion, transformation, migration, and the stage of disease development [4], [5]. To be more precise, the tactile feeling tends to change predominantly as cancer undergoes pathological development and advances from its initial stages to later stages [6]. This valuable information can be considered a pivotal characteristic to determine the stages of CRC polyps.

Colonoscopy is currently the gold standard for assessing CRC as it allows for the direct visualization of the colon’s inner surface (where cancer typically originates), the collection of tissue samples for biopsies, and the treatment of both pre-cancerous conditions and early-stage cancers. However, using this vision-based diagnostic approach, the task of stage detection and classification without sense of touch is highly complex, subjective, and clinician-dependent [7]. Of note, this critical limitation has increased the risk of early detection miss rate and mortality [8]. Notably, these challenges primarily result from the limitations of current colonoscopy technologies that rely on visual observation (e.g., camera occlusions), the absence of tactile feedback, and the methods used to classify polyps. These factors collectively make the early identification of the cancer stage a significant challenge for healthcare providers [9], [10]. These limitations clearly highlight the need for the development of new technologies that directly and safely can provide tactile sensing during a colonoscopic procedure.

To address the aforementioned critical limitations regarding direct force measurement during colonoscopy, and lack of tactile information of the CRC polyps, various instruments with embedded tactile sensors have been developed and reported in the literature. For instance, Chuang et al. [11] introduced a piezoelectric tactile sensor incorporated into rigid endoscopic instruments. Also, Sokhanvar et al. [12] proposed a micro electromechanical-based endoscopic tactile sensor in a layered structure. However, fabrication of this sensor is relatively demanding, and it lacks the capability to effectively capture the force of the surface and the interacted object. Ly et al. [13] also presented a palpation system consisting of a pneumatic tactile display and a force sensor to perceive the tumor size. The reported accuracy of this sensor is 65% for the size which may not be sufficiently sensitive and reliable for an early diagnosis of CRC polyps. A finger-type tactile sensor designed by Arian et al. [14] for the purpose of palpating polyps and diagnosing their hardness. However, this sensor necessitates a high interaction force, which could potentially harm the soft tissue during the diagnostic process. A detailed review of similar technologies can also be found in [15], [16].

As our main contributions and to collectively address the limitations of the existing tactile sensors developed for CRC early diagnosis via a colonoscopic procedure, in this paper, we propose design, fabrication, and characterization of a unique soft and inflatable strain-based tactile sensing balloon (SI-STSB). As shown in Fig. 1, the proposed tactile sensing balloon features the use of a soft and stretchable strain-based sensing gel layer integrated with an inflatable balloon mechanism. Inspired by the design of wearable soft sensors (e.g., [17]–[19]), the strain-based sensor utilizes a liquid metal (i.e., eutectic Indium Gallium alloy, a.k.a. EGaIn) injected within spiral-shape microchannels that are embedded inside a stretchable silicon substrate. When stretched due to the inflation of the balloon mechanism and/or interaction with the environment, the resistance of the injected liquid metal changes due to the change in the geometry of the microchannels. Given two known inputs (i.e., measured resistance of spiral channel and the input pressure of the balloon), the interaction force can be obtained using the introduced one-to-one calibration mapping. The maximum load capacity, hysteresis, and sensitivity of this sensor can also be adjusted by the input pressure to estimate the interaction force, whereas all similar strain-based tactile sensors in the literature (e.g., [17]–[19]) have fixed load capacity and sensitivity with high hysteresis.

Fig. 1.

(a) Conceptual illustration of integrating two SI-STSBs with a colonoscope for direct force measurement of CRC polyps. (b) A closeup view of inflated SI-STSB 1 and deflated SI-STSB 2.

Thanks to its unique structure and features, SI-STSB offers numerous advantages that makes it well-suited for the early-diagnosis of CRC polyps and particularly providing tactile sensing; (i) First, it provides a direct interaction force between the sensor and the polyp that can be used for identification of the polyps and cancer stage [20]. Moreover, this force ensures a safe interaction between the colonoscope and the colon’s surface during the procedure preventing potential tissue damage; (ii) It can readily be integrated with the distal end of existing colonoscopic devices and robots to provide clinicians with additional tactile information while not changing their current surgical workflow and visual feedback; and (iii) Thanks to its inflatable and soft structure, its geometry can be adjusted and adapted to the internal anatomy and variable-size of the colon, allowing us to safely measure interaction forces without altering the existing diagnosis workflow. To demonstrate the features and performance of the SI-STSB, we first fabricated a sensor that can be integrated with a colonoscope. Next, we proposed and experimentally evaluated a unique empirical calibration process and one-to-one analytical mapping for this soft and inflatable sensor.

II. Methodology

A. Working Principle and Diagnosis Procedure

As demonstrated in Fig. 1, the working principle of SI-STSB is very simple yet highly intuitive in which after integration of the sensor with a colonoscope, the interaction force between the inflated sensor and colon environment or polyps can directly be captured by measuring the change in the resistivity of the liquid metal injected inside the microchannels of the sensing gel layer at a known internal pressure of the balloon. Thus, in the envisioned diagnosis procedure, as illustrated conceptually in Fig. 1, the initial step involves integrating one or multiple SI-STSBs (in their deflated mode) with the colonoscope. Subsequently, during the routine screening procedure, as the clinician inserts and advances the colonoscope within the colon, the SI-STSBs can be inflated when necessary or sequentially on their demand. This inflation process is carried out to measure and record the direct interaction forces between the colonoscope and the surrounding environment while ensuring a secure interaction force (i.e., <13.5 N [21]), and a safe pressure level (i.e., <7.6 kPa [22]). It is worth emphasizing that the proposed integration of the proposed sensor with the colonoscope has been inspired by FDA-approved colonoscopic rings and cuffs [23], thus would not cause potential safety complications inside the colon.

B. Design of the SI-STSB

As depicted in Figure 2, the suggested SI-STSB design consists of the following key elements :

Fig. 2.

Proposed design for the SI-STSB. (a) Ring-shape Frame, (b) Torus-shape balloon, (c) Top view of the SI-STSB, (d) Strain-based sensing gel layer.

1). Ring-Shape Frame:

This frame houses (1) a torus-shape balloon that is fixated to the frame, (2) air pressure connections for inflation and deflation of the balloon, and (3) two peripheral notches for fixing the balloon to the frame and preventing the air leakage. The dimensions of the frame, specifically the inner diameter (ID_Ring) and outer diameter (OD_Ring), are determined by considering the outer diameter of the colonoscope, which typically falls within the range of 12–15 mm, and the diameter of the colon itself, which varies between 3 to 8 cm [24]. Also, L_Ring of the ring frame determines the diagnostic area of the colon and is designed based on the size of the balloon, and the area of the strain-based sensing gel layer. Taking into account these design parameters and our utilized PENTAX EC 3840 L colonoscope with 13 mm OD, we selected (i) ID_Ring= 13.2 mm, (ii) OD_Ring= 29 mm, and (iii) L_Ring= 41.5 mm to avoid blocking the colon and creating a potential injury to the colon surface. To minimize the overall size of the ring-shape frame air channels are embedded into the ring and fabricated using high-resolution additive manufacturing (Formlabs Form 3, Formlabs Inc). Fig. 2 shows the fabricated ring-shape frame, assembled pressure lines, and the constructed balloon.

2). Stretchable Torus-Shape Balloon:

The balloon mechanism ensures a safe interaction between the gel layer of the SI-STSB and CRC polyps to enable a direct interaction force measurement. It is made of a stretchable silicone and will be fixed to the ring-shaped frame. Both the inflation and deflation modes of the balloon can be controlled by using a pressure regulator (i.e., Programmable Air), and the required pressure can be provided through air channels incorporated within the attachment ring.

3). SI-STSB Strain-based Sensing Gel Layer:

The sensing gel layer is the most essential component of the SI-STSB that acts as the sensing component of the proposed sensor. Of note, thickness, stiffness, and the fabrication procedure of this layer directly affect the sensitivity and performance of the sensor. As conceptually shown in Fig. 3 (b)–(d), this gel layer uses an EGaIn injected within spiral-shape microchannels that are embedded inside a stretchable silicon substrate. When stretched due to the inflation of the balloon mechanism and/or interaction with the environment, the resistance of the injected liquid metal changes due to the change in the geometry of the microchannels that are used for measurement of interaction forces [25], [26]. Of note, the geometry of the spiral channel directly affects the sensitivity, load capacity, and hysteresis behavior of the sensing layer. For this paper, without loss of generality and considering the mentioned design and size limitations of the colon geometry, we have selected a spiral-shape microchannel geometry with a circular cross section and the following critical dimensions shown in Fig. 2 (d): radius of microchannel cross section r_Channel= 0.55 mm, maximum coil diameter D_Coil= 24.5 mm, and Coil gap c= 1.1 mm. More details about the geometry of this spiral pattern can be found in [25], [26]. Moreover, L_Gel and H_Gel are determined as 42.5 mm to easily place the sensing layer on the stretchable torus-shape balloon. Of note, the SI-STSB can incorporate separate sensing patterns that are specifically designed for different sizes with varying force thresholds and precision requirements. By adjusting the length and cross-section of the sensing patterns within the flexible balloon, this sensor can be customized to suit multiple environments, resulting in enhanced measurement of interaction forces.

Fig. 3.

Two-step fabrication procedure for the proposed SI-STSB components.(a) Designed 3-part mold for fabrication of the torus-shape balloon, (b) Assembly of the mold for the silicone mixture pouring, (c) Fabricated torus-shape stretchable balloon after pouring and curing the silicone mixture, (d) Torus-shape stretchable balloon prototype, (e) Top: 3D printed inner spiral microchannel next to the US quarter for size comparison; Bottom: Example prototype of the fabricated strain based sensing gel layer, (f) Designed mold for fabrication of the sensing gel layer, in which silicone mixture is poured into as the first layer, (g) 3D printed spiral pattern is placed on the cured silicone mixture, and the second layer of silicone mixture is poured on top, (h) Finalized sensor is removed from the mold and placed into the acetone to dissolve the embedded spiral model, (i) After dissolving, EGaIn is injected into the created microchannels, and (j) SI-STSB prototype after attaching both balloon and the sensing layer together.

C. Fabrication of the Inflatable Balloon and Strain-based Sensing Layer

Fig. 3 describes a two-step molding procedure for the fabrication of the inflatable balloon and the sensing gel layer in which we first fabricate the inflatable balloon, then, fabricate the sensing gel layer with embedded channels on this substrate, and finally adhere them together.

1). Inflatable Balloon:

As shown in Fig. 3 (a)–(c), to fabricate an inflatable torus-shaped balloon for the proposed SI-STSB, our initial step involved the design and additive manufacturing of a cylindrical mold comprising three components: an inner circular rod (a1), a hollow tube (a2), and a bottom fixture (a3) specifically designed for the rod. We utilized an FDM 3D printer (Raise 3D Pro2 Series, RAISE3D Inc.) and Art White PLA filament (RAISE3D Inc.) to build the cylindrical mold. The hollow tube has an inner diameter of 26 mm, while the outer diameter of the inner circular rod measured 24 mm. The total length of the completed mold is 60 mm. Subsequently, to build the inflatable balloon using this manufactured mold, we employed a soft, biocompatible platinum-cure silicone known as Ecoflex 00–30 (Smooth-On, Inc.), mixing Part A and Part B, in a precise 1:1 ratio. This silicone provides moderate deformability and shore hardness (200 psi tensile strength), as noted in [27], [28]. Prior to initiating the molding process, we coated the surface of the three-part mold with Ease 200 (Mann Technologies) to prevent any unwanted adhesion. Following this preparatory step, the silicone mixture was placed into a vacuum chamber to effectively eliminate any trapped air bubbles within it. Once the degassing process was completed, the silicone mixture was carefully poured into the three-part mold, filling the empty space between the hollow tube and the inner rod. After allowing the mixture to undergo a curing process for approximately 4–5 hours at room temperature, the mold was disassembled to obtain the fabricated durable, soft, and inflatable balloon.

2). SI-STSB Strain-based Sensing Gel Layer:

To fabricate a strain-based sensing gel layer (length × width × thickness as 42.5 mm × 42.5 mm × 4 mm), we used Ecoflex 00–30 silicon (Smooth-On, Inc.) and followed the sacrificing method proposed in [29]. We first constructed a single rectangular mold (70 mm × 52.5 mm × 8 mm) including a rectangle opening (60 mm × 42.5 mm × 4 mm) in the middle for pouring of the silicone mixture of the first layer of the sensor after degassing process in a vacuum chamber (Fig. 3 (f)). Notably, the dimensions of the sensing gel layer of SI-STSB are determined by the size of this opening. After pouring the degassed silicone mixture into the rectangular opening, and curing for 45 minutes at 75°C, we embedded the internal circular pattern printed with an FDM 3D printer (Raise 3D Pro2 Series, RAISE3D Inc.) and ABS filament (RAISE3D Inc.). Next, we again poured the degassed silicone mixture into the rectangular opening to cover the embedded circular pattern (Fig. 3 (g)). After curing, we put the fabricated sensor layer into the acetone to dissolve the ABS filament and obtain the microchannel (Fig. 3 (h)). As the final step, EGaIn (MSE Supplies) was injected into the microchannels of the SI-STSB and sealed (Fig. 3(i)). Finally, both the balloon and the obtained sensing gel layer were adhered together with silicone rubber adhesive, Sil-Poxy (Smooth-On, Inc.).

III. Evaluation Experiments and Results

A. Deformation Analysis of the Utilized Silicone

To ensure safe application of the fabricated inflatable balloon inside the colon, we first need to analyze the deformation behavior of the utilized silicone (i.e., Ecoflex 00–30) and obtain its maximum elongation before failure. In this regard, we first fabricated three dog-bone-shaped specimens following the recommended geometries (shown in Fig. 5) by the ASTM D412 type C standard and the literature [30]. As shown in Fig. 4, we then used the Instron Tensile machine (Instron, Massachusetts, USA) to conduct experiments on either of the specimens. In particular, through these experiments, we found the stress-strain relationship and the maximum elongation before failure for each specimen. Figure 5 shows the obtained results of these three experiments. Using the obtained results, we then calculated the Young’s modulus of the silicon samples (i.e., σ = 0.03ϵ and E = 30kPa) – as the slope of the linear section of these plots corresponding to our operational region – as well as the average 1024% elongation before failure. Of note, our results are in good agreement with the literature (e.g., [27], [28]) and the data sheet provided by the manufacturing company (Smooth-On, Inc.). Considering the 1024% maximum elongation failure of the used silicon, we can ensure a safe operation of the balloon inside the colon’s cavity.

Fig. 5.

Figure depicts the stress-strain curve for the uni-axial tension test, the maximum elongation before failure, along with the rupture location of each dog-bone specimen. The figure also includes a zoomed-in view of the initial part of the curve, with a linear fit used to obtain the modulus of elasticity in our operational region.

Fig. 4.

Experimental setup used to obtain the deformation behavior of the used silicone:①- single column Instron Tensile Tester 68TM-5 (Instron, Massachusetts, USA), ②- Stretched dog-bone specimen, ③- Fixed tensile grip, ④- Moving tensile grip, ⑤- Static load cell with 5 kN capacity, and ⑥-Screen and software UI for plotting the strain-stress curve. Figure also shows the geometry of the fabricated dog-bone-shaped silicone specimen following ASTM D412 type C standard.

B. Experimental Setup

Fig. 6 demonstrates the experimental setup used for characterization and calibration of the fabricated SI-STSB. Particularly, we used this setup to obtain the relationship among the applied force F, balloon’s internal Pressure P, and the variable resistance R of EGaIn. Using this relationship, in the envisioned colonoscopy application, given known balloon pressure and resistance of the EGaIn during interaction, we can directly estimate the interaction forces between the sensor and the CRC polyp. As shown, the proposed setup consists of the fabricated SI-STSB, a digital force gauge with 0.02 N resolution (Mark-10 Series 5, Mark-10 Corporation) attached to a single-row linear stage with 1 μm precision (M-UMR12.40, Newport) providing precise movement for the force gauge probe to apply known forces to the sensor’s gel layer, a multimeter (Fluke 117 Digital Multimeter) to measure the resistance changes with 0.1 Ω resolution, a commercial colonoscope (PENTAX EC 3840 L) for the demonstration of the applicability of the proposed sensor. The setup also uses a pneumatic actuation system (Programmable Air, Crowd Supply) for controlling inflation and deflation of the SI-STSB as well as measuring the pressure inside the balloon. This pneumatic actuation system consists of two compressor/vacuum air pumps, three solenoid valves, and an on-board air pressure sensor with 0.1 kPa resolution. We also utilized Arduino Nano as a basic data acquisition and controller to record the pressure inside the balloon and send a command to the compressor/vacuum pump, and solenoid valves for controlling the pressure inside the balloon and preventing leakage.

Fig. 6.

Experimental Setup: ① - Fluke Multimeter with 0.1 ohm precision, ②- soft and inflatable strain-based tactile sensing balloon, ③- Programmable Air, ④- A zoomed view of the SI-STSB and fixture, ⑤- Flat and rigid CRC polyp phantom, ⑥- Mark-10 Series 5 Digital Force Gauge, and ⑦- M-UMR12.40 Precision Linear Stage. Figure also shows a schematic view of the pressure inside the balloon and applied force for three different cases such as (i) free inflation (F = 0, P), (ii) Interaction without inflation (P = 0, F), and (iii) Interaction with inflation (P, F) for the experimental characterization.

C. Experimental Procedure

To perform the calibration procedure, as shown in Fig. 6, SI-STSB was initially secured on a 3D-printed rigid fixture on an optical table. Next, a 3D printed flat CRC polyp phantom was attached to the force gauge using the threaded connection printed at the base of it. After preparing the hardware (i.e., the Arduino Serial Monitor, Multimeter, and Force Gauge) for measurements, we gathered data in three distinct cases for the calibration, described below.

Free Inflation (P ≠ 0 and F = 0 in Fig. 6):

In this set of experiments, The balloon’s internal pressure was gradually increased (i.e., Loading step) and then decreased (i.e., Unloading step) using the Programmable Air setup while there was no interaction force against the balloon’s inflation. Of note, this case allowed us to study the relationship between the change in the EGaIn’s resistance solely due to the changes in the balloon’s internal pressure.

Interaction without Inflation (P = 0 and F ≠ 0 in Fig. 6):

In this case, we kept the balloon at the deflated mode while incrementally increased (i.e., Loading step) and then decreased (i.e., Unloading step) the interaction force applied to the strain sensor using the force gauge and the linear stage mechanism. This scenario helped us to understand how the sensor’s resistance responds to an external force in the absence of the balloon’s inflation.

Interaction with Inflation (P ≠ 0 and F ≠ 0 in Fig. 6):

In this case, we introduced incremental increase (i.e., Loading step) and then decrease (i.e., Unloading step) in the interaction forces to the sensor’s gel layer, starting from zero and progressing to a maximum value (shown in Fig. 8(a)). Through this process, we maintained a constant pressure inside the balloon. We performed this experiment at three different pressure levels (i.e., 2.17 kPa, 2.74 kPa, 3.21 kPa) to observe the sensor’s response under varying conditions (shown in Fig. 8 (a)). Of note, since the diameter of the colon is roughly 60 mm in most parts [31], the selected pressure range for these experiments was designed based on the minimum pressure required for the balloon to contact the fixture (which is the same size as the colon). Moreover, the maximum pressure inside the balloon was limited by the maximum allowable pressure inside the colon (i.e., <7.6 kPa [22]) to ensure safety of procedure. During the procedure, we recorded the pressure level inside the SI-STSB by an on-board pressure sensor of the Programmable Air and Arduino Nano, the resistance of the liquid metal by Multimeter, as well as the synchronized interaction force between the SI-STSB and the objects via the Force Gauge. Our methodology emphasized robust data collection practices, wherein each experiment at each internal pressure was repeated three times across the entire range of interaction forces consecutively to ensure repeatability. In total, we executed 12 complete tests (including 4 pressures and three times of repetitions for each pressure) and collected a total of 480 data for both loading and unloading cycles.

Fig. 8.

(a) Results of the calibration experiments under different internal balloon pressure and interaction forces; (b) Loading and Unloading regression models. The shaded regions represent the hysteresis in each case.

Figure 7(a) depicts the 3D plot of the measured resistances versus two other inputs (i.e, corresponding interaction forces and pressures inside the balloon) as well as their variations across the performed experiments at three aforementioned scenarios. As shown, this figure illustrates the free inflation behavior of the sensor (i.e., P ≠ 0 and F = 0) represented by the orange plane, interaction without inflation (P = 0 and F ≠ 0) represented by the green plane, and interaction with inflation at three different pressures (i.e., P ≠ 0 and F ≠ 0) represented by the red, black, and blue planes. Fig. 7(a), also displays an example situation (illustrated by arrows) in which the sensor first freely inflates without any interaction (i.e., moving on the orange plane with zero interaction force) until it interacts with the environment (e.g., colon’s surface) at a specific pressure (i.e., 2.17 kPa). After interaction, for a fixed internal pressure, the Resistance-Force relationship follows the shown behavior illustrated on the red plane for the loading and unloading phases to reach a specific point. Fig. 7(b) shows the 2D projected view of this 3D plot highlighting a specific case of inflation without interaction force. The orange plane demonstrates the relationship between resistance change and the internal pressure. This view also shows the projected view of the planes of interaction without/with inflation as vertical lines. The other 2D view of this plot is shown in Fig. 8 in which the relationship between interaction force and resistance at each constant pressures have been projected.

Fig. 7.

(a) 3D plot of the measured resistances versus two other inputs (i.e, corresponding interaction forces and pressures inside the balloon) as well as their variations across the performed experiments at three aforementioned scenarios: free inflation (represented by the purple plane), interaction without inflation (represented by the green plane), and interaction with inflation at three different pressures (represented by the red, black, and blue planes). Figure also highlights a point on the graph and its path line indicated by black arrows representing a free inflation scenario followed up by interaction with inflation scenario at the constant 2.17 kPa. (b) Projected view of the 3D plot in free inflation mode for loading (represented by orange line) and unloading (represented by orange dash line) cycles. This view also shows the projected planes of interaction without/with inflation as vertical lines.

D. Loading and Unloading Calibration Functions

In the envisioned colonoscopy procedure using SI-STSB, we are interested in finding the interaction force F between the sensing layer and the colon’s surface given known internal pressure P of the balloon and the measured resistance variable resistance R. Therefore, considering the obtained experimental results, shown in Fig. 8 (a), our main objective in this section is to find a unique analytical function 𝒢 that relates the interaction force F and the internal Pressure P of the balloon to the variable resistance R (i.e., F = 𝒢(R, P)). To find the function 𝒢, we first need to find a calibration function 𝒞 that relates the measured resistances in the performed experiments to the applied internal pressures and interaction forces (i.e., R = (F, P)). Knowing 𝒞, we can then obtain the desired G function as follows: 𝒢 = 𝒞⁻¹(F, P).

Based on the measured data shown in Fig. 8 (a), we opted to use the vertex form of a quadratic equation to model the calibration function C during the Loading step of the performed experiments. This particular form of the quadratic equation (i.e., y = a(x − h)² + k) provides a clear representation of the vertex point (h, k) in which parameter h indicates the horizontal translation along the x axis and parameter k is the vertical translation along the y axis. Moreover, coefficient a defines the curvature or steepness of the quadratic function. of note, either of these parameters is variable and depends on the balloon’s pressure:

$$R_L=a(P\text{)(}F-h(P{\text{))}}^{\text{2}}+k\left(P\right)$$ (1)

where R_L indicates the resistance in the loading step.

Moreover, the investigation of the unloading cycle in Fig. 8 (a) demonstrates a linear relationship for the performed experiments except for the free inflation case. In this case, a noteworthy distinction arises, where both the loading and unloading cycles manifest as quadratic relationships. To facilitate a quick and straightforward comparison between the interaction without inflation and the interaction with inflation versions, a linear approximation has been employed for characterizing the unloading cycle of the interaction without inflation:

$$R_{UL}=a\left(P\right)F+b\left(P\right)$$ (2)

where R_UL indicates the resistance in the Unloading step.

By applying regression analysis (fit function in MATLAB), we successfully determined the optimal values of parameters a, h, and k that minimized the discrepancies between our observed data points and the predicted values derived from (1) and (2). Fig. 8 (b) shows the fit functions on either of the Loading (cycle B/b - C/c) and Unloading (cycle C-D or d-e) steps of the performed calibration modes. Also, Table I summarizes the obtained coefficients for either of these experiments.

TABLE I.

Calibration Coefficients of Loading and Unloading Models

Pressure	Loading Coefficients			Unloading Coefficients		Hysteresis
	a	h	k	a	b	Area (N.Ω)	Relative
P0 = 0 kPa	0.09	1.61	0.13	0.59	−1.25	2.1	1.00
P1 = 2.17 kPa	0.37	1.89	0.26	1.04	−2.15	0.33	0.16
P2 = 2.74 kPa	0.57	2.78	0.48	1.54	−4.59	0.38	0.18
P3 = 3.21 kPa	0.32	2.97	0.57	1.10	−3.33	0.39	0.19

After obtaining the coefficients for the Loading cycle, we can calculate the inverse function 𝒢 to estimate the force F* using the measured pressure P and the read resistance R. Using (1), the force estimation equation can be calculated as follows:

$$F*=\sqrt{\frac{R_L-k\left(P\right)}{a\left(P\right)}}+h(P)$$ (3)

Figure 9 illustrates the obtained analytical inverse functions calculated for either of the three considered pressures.

Fig. 9.

Figure compares the estimated interaction force F* at different balloon pressures with the considered validation dataset.

E. Evaluation Experiments

To evaluate the obtained F* inverse functions, we also performed three different experiments at two different pressures of 2.17 kPa and 3.21 kPa. Of note, these data were not used during the regression and calibration steps. For these experiments, we input the known resistances and pressures to (3) and calculated the estimated force F*, and compared it with the measured interaction force obtained through experiments. Fig. 9 compares the estimated values F* with the measured data, demonstrating the high accuracy of the proposed empirical model with a root mean square error of 0.08 (RMSE = 0.08 N) and a maximum absolute percentage error of less than 3% (MPE<3%). The RMSE, MPE, maximum error (ME), mean absolute error (MAE), R-squared score (coefficient of determination), mean error (μ), and standard deviation of error (σ) for three validation sets are represented in Table. II. Figure 10 also illustrates the percentage error for all validation sets in their actual force measurements.

Fig. 10.

The plot shows the percentage error for all validation sets in their actual force measurements.

To represent the hysteresis of the fabricated sensor under different pressures, we also calculated the area between the Loading and Unloading curves (shown in Fig. 8 (b)). Of note, a perfect sensor should show a zero area between the Unloading and Loading modes. Table I summarizes the obtained areas for either of the performed experiments. To demonstrate the integration of the SI-STSB with the colonoscope and its performance within a constrained environment of the colon, we also performed a simulated colonoscopic procedure inside a U-shape phantom made of rigid transparent pipes. Fig. 11 illustrates the functionality and inflation and deflation steps of the sensor inside this phantom.

Fig. 11.

Evaluation of the integrated SI-STSB with a commercialized colonoscope inside a U-shaped pipe in the deflated (a) and inflated (c) modes. Figure also shows a close view of deflated (b) and inflated (d) modes of the SI-STSB integrated with the colonoscope.

IV. Discussion and Conclusion

In Fig. 7, every color-coded plane represents the resistance-interacting force relationship of the sensor in a constant internal balloon’s pressure. Also, for a specific case of inflation without interaction force (shown in Fig. 7(b)), the orange plane demonstrates the relationship between resistance change and the internal pressure alone. This 3D plot clearly demonstrates how the mentioned three parameters (i.e., internal pressure, interacting force, and resistance) are affecting the sensor performance and also how by moving from one internal pressure to another one, the performance of the sensor in interaction with external forces changes. Moreover, this plot distinctly illustrates the key features of our inflatable sensor in which the inflation of the proposed sensor not only enhances its sensitivity (defined by the slope of the loading and unloading forceresistance plots at every pressure) but also reduces hysteresis (defined by the area of the loading and unloading plots) in the sensor compared with the existing flat sensors in the literature. We can clearly see that as we increase the pressure, the slopes of loading and unloading plots are changing while the area between them is reducing. In other words, we can change the performance of the sensor by changing the internal pressure of the balloon while having a unique correlation between the involved parameters. These important features, have also been demonstrated in the projected views showing the relationships between force and resistance at every pressure in Fig. 8 and Table I.

We should also emphasize that the shown green plane in Fig. 7(a) represents the performance of our sensor in an uninflated mode (i.e., internal Pressure P = 0 kPa) that is identical to the performance of the existing non-inflatable strain sensors of the literature (e.g., [25], [26]). In other words, if we do not inflate our sensor, it performs similar to the existing non-inflatable strain sensors presented in the literature that display a high hysteresis (i.e., larger area between the loading and unloading phases) and low sensitivity (i.e., smaller slope of the loading unloading phases) outputs due to the mechanical properties of the utilized silicon substrate [17]. Nevertheless, as can be seen in Fig. 7, Fig. 8, and the calculated hysteresis area in Table I, by increasing the internal pressure, we can enhance the sensitivity (i.e., higher a coefficient) and reduce hysteresis (about six times less) compared with the existing sensors. This performance can be correlated to our innovative approach to measure interaction forces by integrating a flexible strain sensor with an inflatable balloon to introduce controlled pre-tension on the strain sensor. Because of this pre-tension, most of the applied forces deform the sensor’s cross-sectional area rather than generating internal stress within the silicon substrate, thereby improving the sensitivity of sensor. Therefore, our proposed sensor features a variable load capacity and sensitivity that can be adjusted by the internal pressure of the balloon. This adjustment provides a broader range of load capacity and sensitivity for accurately measuring interaction forces, in contrast to the fixed load capacity and sensitivity of existing strain-based sensors proposed in the literature (e.g., [25], [26]).

By utilizing the vertex form of the quadratic function as the calibration function, shown in (1), we could precisely model the curvature and characteristics of the Loading cycle across different pressures and interaction forces. Specifically, parameter a represents the sensitivity of resistance changes in response to exerted force, h signifies the presence of a dead zone (section A/a - B/b in Fig. 8 (b)) in our sensor (i.e., the minimum interaction force that we see a resistance change), and k represents the minimum sensible force by the fabricated sensor. These findings enhance our understanding of the variable relationships between these parameters and the balloon’s internal pressure, maximum load capacity, hysteresis, and sensitivity. Based on Table I, it is evident that the sensitivity of the inflatable modes (defined by coefficient a) has at least doubled compared to the performed deflated experiment. This suggests that the inflatable version is more responsive to changes in the interaction load making it a perfect candidate for polyp characterization through a tactile interaction. Figure 10 and Table II highlights the accuracy and reliability of our proposed calibration method in estimating the force values F* compared with their actual values using the known pressure and resistance measurements. Fig. 11 also shows the functionality and morphability of the proposed sensor in the confined areas (e.g., a colon) that clearly demonstrates safe interaction of this sensor with the anatomy.

TABLE II.

Calculated Errors between estimated force values F* using pressure and resistance compared to actual force values.

set	ME (N)	MPE (%)	MAE (N)	RMSE (N)	R2	μ (N)	σ (N)
1	0.11	3.0	0.03	0.04	0.989	−0.008	0.046
2	0.24	1.7	0.06	0.08	0.976	−0.045	0.028
3	0.22	2.0	0.05	0.07	0.977	−0.038	0.023

To conclude, in this study, with the goal of enhancing the early diagnosis of CRC polyps using the current colonoscopic procedures and providing the tactile feedback to the clinicians, we introduced SI-STSB. SI-STSB can seamlessly be integrated with the existing colonoscopic devices without disrupting the current diagnosis workflow. The maximum load capacity, hysteresis, and sensitivity of this sensor can also be adjusted through the input pressure to accurately estimate the interaction force. To fabricate this sensor, we proposed a unique fabrication procedure enabling to readily scale the size of the fabricated spiral-shape channels within the silicon substrate of SI-STSB. We also proposed and evaluated a calibration procedure together with a complementary empirical calibration function for this sensor. Results demonstrate high accuracy (RMSE= 0.07 N and MAE < 3%) of the proposed calibration function compared with the evaluation experiments.

Despite the demonstrated features of the proposed stretchable and inflatable sensor, it only has a localized and discrete sensing area with asymmetric inflation behavior. In the future, we plan to fabricate a balloon mechanism with a fully covered sensing layer that introduces symmetric inflation. Moreover, we will perform user studies and test this sensor in a more realistic surgical setting and on ex-vivo animal models. Also, as our long-term goal, after approval of IRB protocol and FDA clearance, we also plan to perform in vivo animal and human clinical studies.

Biographies

Mohammad Rafiee Javazm received his B.Sc. and M.Sc. degrees in Mechanical Engineering from Amirkabir University of Technology and Sharif University of Technology, Tehran, Iran, in 2016 and 2018, respectively. Currently, he is pursuing his Ph.D. in surgical robotics at the Advanced Robotic Technologies for Surgery Laboratory, Texas Robotics, Walker Department of Mechanical Engineering, University of Texas at Austin. His research focuses on designing and developing robots for medical applications, with interests in soft robotics and dynamics and control.

Ozdemir Can Kara received his B.S. and M.S. degrees in Mechatronics Engineering from Sabanci University, Istanbul, Turkey, in 2015 and 2018, respectively. From 2018 to 2021, he was a Research Assistant with the Reconfigurable Robotics Laboratory, École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland. He is currently working toward the Ph.D. degree in surgical robotics at the Advanced Robotic Technologies for Surgery Laboratory, Texas Robotics, Walker Department of Mechanical Engineering, University of Texas at Austin. His research interests include the design and development of a framework consisting of soft sensors and machine learning for the early detection of colorectal cancer, soft robotics, compliant mechanisms, haptic interfaces, actuation, and sensing for medical applications.

Farshid Alambeigi (Member, IEEE) received the B.Sc. and M.Sc. degrees in mechanical engineering from the K.N. Toosi University of Technology and the Sharif University of Technology, Tehran, Iran, in 2009 and 2012, respectively, and the M.S.E. degree in robotics and the Ph.D. degree in mechanical engineering from the Johns Hopkins University, Baltimore, MD, USA, in 2017 and 2019, respectively. He is currently an Assistant Professor with the Walker Department of Mechanical Engineering and the Texas Robotics at the University of Texas at Austin, TX, USA. His research interests include development, sensing, and control of high-dexterity continuum manipulators, soft robots, and flexible instruments and sensors designed for less-invasive treatment of various medical applications.