Biosafety-cabinet wipe-cleaning assessment system: Integration of visual, thermal, and pressure sensing

Abstract

Introduction

Reliable surface disinfection of biosafety cabinets is essential to prevent contamination in pharmaceutical, clinical, and laboratory environments. However, manual cleaning using ethanol (ETH) and wiping is subject to operator variability, making objective verification of disinfection quality difficult. This paper proposes an integrated disinfection-validation system that combines thermal imaging, visual tracking, and pressure sensing to objectively assess the adequacy of ETH spraying and pressure-based wiping.

Methods

The proposed system integrated three sensing modalities: Red-green-blue (RGB) imaging, thermal sensing, and pressure detection. The RGB and thermal cameras (placed outside the biosafety cabinet) monitored the stainless steel (SUS) work surface, and the flexible pressure sensors (embedded in the wiping tool) detected the applied force. The system was biologically validated using Bacillus subtilis spores on SUS plates, using ETH spraying only, wiping only, and combined spraying and pressure-based wiping. Reproducibility was evaluated under different wiping speeds and operator techniques, using time-series and endpoint analyses.

Results

The system successfully visualized the sprayed, wiped, and untreated areas in real time. Only the combined treatment resulted in complete inhibition of bacterial growth, demonstrating the synergistic effect of chemical and mechanical actions in achieving effective disinfection. The classification accuracy and quantitative detection of the cleaned areas were consistent across the trials, confirming their reproducibility under variable wiping patterns.

Conclusions

This integrated sensing system provides a reliable and objective method for validating surface disinfection in biosafety cabinets. By enabling the real-time, recordable, and quantitative assessment of manual cleaning, it bridges the gap between procedural operations and verifiable decontamination, supporting standardized and data-driven hygiene management in biosafety and cell processing environments.

Highlights

• •Visual, thermal, and pressure sensing, integrated for real-time cleaning validation.
• •Integrated system for evaluating wiping-based disinfection in biosafety cabinets.
• •Objective, recordable, and quantitative assessment of manual wiping enabled.
• •System reproducible across repeated trials and variable wiping patterns.
• •Effective removal of surface bacteria by ethanol spraying plus wiping.

1. Introduction

Ensuring the sterility of biosafety cabinet (BSC) work surfaces is an important requirement in cell-product manufacturing, where terminal sterilization of the final product is often infeasible owing to the biological materials involved [[1], [2], [3], [4]]. Accordingly, strict contamination-control procedures, (including ethanol (ETH) spraying, wiping with nonwoven fabric, and ultraviolet (UV) irradiation) are implemented routinely [[5], [6], [7], [8], [9]]. However, the wiping process (though widely used) remains highly operator dependent and lacks objective evaluation criteria. For example, microbiological assessment requires several days of incubation on agar plates, making it impractical to wait for the results before initiating subsequent production steps. Consequently, operators must proceed without timely confirmation of cleaning adequacy and often remain uncertain whether they have applied sufficient pressure or adequately covered the work surface [10,11]. Moreover, contaminants such as endotoxins, proteins, and DNA derived from raw materials and culture media cannot be effectively removed by ETH or UV light alone; mechanical wiping is required for physical removal [7,9]. These limitations highlight the need for a method that not only quantifies the wiping quality but also provides immediate visibility during the cleaning process.

Recent advances in sensing technologies have facilitated the objective monitoring of manual tasks in controlled environments. Thermal cameras can visualize ETH spraying by detecting the transient cooling of work surfaces, RGB cameras can track wiping trajectories through image analysis, and pressure sensors can quantify the mechanical force applied during cleaning. Prior studies have shown that manual cleaning results in highly variable biofilm removal from complex surfaces, emphasizing the difficulty in achieving consistent decontamination using conventional approaches [12]. Thermal imaging has also been used to visualize human-contact points on environmental surfaces, suggesting its potential utility in disinfection planning [13]. However, each sensing modality provides only a single dimension of information (spatial coverage, fluid distribution, or mechanical pressure) and cannot independently determine whether cleaning is performed thoroughly, in the correct sequence, or before ETH evaporation. Furthermore, existing systems generally lack real-time integrated assessments, which limit their ability to support either immediate decision-making or operator training. Together, these limitations underscore the need for an integrated real-time monitoring framework capable of comprehensively evaluating the surface-disinfection performance of BSCs.

To address these challenges, we developed an integrated system for the assessment of wipe cleaning that synchronizes RGB imaging, thermal detection, and pressure sensing to achieve a real-time recordable evaluation of wiping-based decontamination. The system was designed to operate noninvasively without attaching sensors to the operators or modifying the BSC structure.

We first validated its biological performance using Bacillus subtilis as a model organism and then assessed its temporal characteristics and reproducibility under various wiping conditions. This approach aimed to support safer and more standardized practices in cell-processing environments by enabling the quantitative, reproducible, and real-time assessment of cleaning operations.

2. Materials and methods

2.1. Integrated wipe-cleaning assessment system: configuration

We developed an integrated wipe-cleaning assessment system that evaluated wipe-cleaning procedures in BSCs using synchronized RGB imaging, thermal imaging, and pressure sensing (Fig. 1A, B, and 1C). The system was designed to operate noninvasively from outside the BSC without requiring any wearable devices for operators.

Fig. 1.

Integrated wipe-cleaning assessment system: Overview.

(A) Overview of integrated wipe-cleaning assessment system. The top panel shows a photograph of the complete system. The lower left shows the device structure, and the lower right indicates the placement of the thermal and visual cameras. (B) Signal flow, including data acquisition, processing, and judgment steps for evaluating ETH spraying, wiping coverage, and pressure application. (C) Pressure sensing module. Eight pressure sensors are embedded in the wiping pad. The top panel shows the sensor layout, and the middle panel shows representative visual states when the pressure is Off and On. The lower panel displays a representative pressure-sensor output trace with the predefined threshold, the approximate 1 kg reference line, and the regression equation derived from the load-calibration experiment, providing a practical interpretation of the sensor signal during wiping.

The RGB camera (GZ320, Gazo, Niigata, Japan) unit was positioned to observe the work surface and capture the wiping trajectory in real time. It traced the position of a color marker, attached to the wiping implement. To avoid the interference of reflections on the stainless steel (SUS) surface, three colors with different reflection attributes were selected as markers. These colors were easily distinguishable in the hue-saturation-value color space. The acquired images were processed to track the position of the wiping implement and generate an orthographic projection image. The orthographic projection was achieved using a projective transformation, which mapped one quadrilateral region to another. To achieve accurate spatial alignment between the RGB and thermal images, we prepared a custom-made calibration board consisting of a glass-epoxy substrate (25 cm × 25 cm) on which four reference points were arranged. Each reference point comprised a heated resistor detectable in the thermal image, and a light-emitting diode placed in close proximity detectable in the RGB image. Prior to each experiment, the calibration board was captured after the resistors had reached a stable temperature, and the four reference points were extracted from each image. Next, the projective transformation matrix was calculated using these four-point correspondences, enabling the RGB image to be mapped onto the same coordinate space as the thermal image for subsequent integrated analysis. A spatial map of the wiping coverage was then plotted, and the wiped positions were computed from the projected image.

In the present study, thermal imaging (Boson + AS50deg, Teledyne FLIR OEM, OR, USA) detected the presence and distribution of the disinfectant—by capturing the transient temperature decrease on the SUS work surface—caused by spraying a 70% ETH disinfectant solution (76.9–81.4% w/v ETH and 3.7% w/v isopropyl alcohol; Yamazen Pharm Co., Osaka, Japan). A background image was captured before disinfection began, and image subtraction was performed with reference to it to generate a thermal difference image. When a disinfectant is sprayed, it first drifts through the air, causing fluctuations in the captured temperature. When the disinfectant lands on SUS surfaces, the surface temperature drops owing to the latent heat of vaporization. The disinfected area was detected as follows. After computing the thermal difference image by subtracting the pre-experiment baseline image, a noise reduction procedure was applied: a 3×3 median filter, binary thresholding, and one cycle each of erosion and dilation using a 3×3 kernel. ETH-sprayed regions were then detected using a two-stage criterion: pixels where the temperature had decreased by ≥ 1.5 °C from the baseline image were designated as candidate regions. Among these regions, only pixels where the frame-to-frame temperature change remained lower than 1.5 °C for five consecutive frames (approximately 700–800 ms at imaging rates of 6–7 frames per second) were classified as confirmed sprayed regions, indicating stable post-spray surface cooling owing to ETH evaporation.

The pressure-sensor unit (Fig. 1C) was a flexible sheet-type pressure sensor (FSR03DE, Ohmite, IL, USA) attached to the wiping surface of the cleaning tool (ZQ21-2014, KOHNAN SHOJI Co. Ltd., Osaka, Japan). During wiping, a BEMCOT wipe (Asahi Kasei Co., Tokyo, Japan) was attached to the sensors; to avoid impairing the performance of the cleaning tools, thin and flexible sensors were used. The pressure-sensor inputs from the eight channels are represented by $v_1,\dots ,v_8$; and the baseline value with no pressure is $c$. The deviation score ($V$) is calculated as follows.

$$\begin{array}{c}V=\sum _{i=1}^8{\left(v_i-c\right)}^2\end{array}$$

The raw electrical-resistance values were used as continuous pressure signals in subsequent analysis.

A calibration experiment was performed using a weighing scale to provide an approximate physical interpretation of the pressure-sensor output. Known loads were applied to the wiping device, and the averaged offset-corrected output across the eight sensor channels was recorded. Next, a calibration curve was generated using a power-law to describe the relationship between the sensor output and applied load (kg). The resulting regression equation was used to estimate the approximate load corresponding to the sensor signal during wiping (Fig. 1C, lower right).

2.2. Evaluation of wipe-cleaning using bacterial surface contamination

To evaluate the biological performance of the system, a surface contamination model was constructed. Bacillus subtilis was the biological indicator (NBRC 13722, Bioball Multishot 10E8; bioMérieux, Marcy l'Etoile, France) used; it was dissolved in a rehydration fluid (bioMérieux) and diluted to 1 × 10³ colony-forming units (CFUs) in physiological saline (Otsuka Pharmaceutical Co. Ltd., Tokushima, Japan). SUS plates (SUS304, 5 cm × 5 cm; AS ONE Co., Osaka, Japan) were inoculated with 200 μL of a saline suspension and left to air-dry under sterile conditions (Fig. 2A).

Fig. 2.

Integrated wipe-cleaning assessment system: Biological-disinfection efficacy.

(A) Overview of the biological assay. Stainless steel (SUS) plates inoculated with Bacillus subtilis (B. subtilis) were air-dried and subjected to different cleaning conditions. Wipe-cleaning effectiveness was evaluated. CFU indicates “colony forming units”. (B) Representative images of soybean casein digest (SCD) agar plate showing residual B. subtilis colonies after three treatment conditions: disinfectant only, wiping only, and combined treatment including disinfectant and wiping. (C) Quantification of residual colonies cultured on SCD agar. Asterisks (*) indicates p < 0.05, as determined by the Kruskal–Wallis test followed by Dunn's multiple comparisons test. ND indicates "not detected"; LOD indicates "limit of detection."

The inoculated plates were subjected to wiping treatments under three different conditions:

• (1)ETH spraying only (no wiping pressure),
• (2)wiping only (without ethanol), and
• (3)combined treatment with both ethanol spraying and pressure-based wiping.

The experimental setup and evaluation process are illustrated in Fig. 2A.

After treatment, B. subtilis were recovered from each SUS plate by pipetting 200 μL of sterile physiological saline onto the plate surface and scraping the liquid across the surface using a micropipette tip, following the methodology described in our previous study [7]. The resulting suspension was aspirated and transferred to a tube. The collected solution was diluted, seeded on soybean casein digest agar medium (Shimadzu Diagnostics, Tokyo, Japan), and incubated at 37 °C for 24–48 h. The resulting colonies were photographed, and the CFU number was counted for quantitative comparison.

2.3. Integrated wipe-cleaning assessment system: timeline-based and endpoint evaluation

Timeline-based evaluation was performed by manual zigzag wiping motions in a BSC while simultaneously recording RGB videos, thermal images, thermal-difference maps, pressure signals, and classified outputs. Four annotated time points were defined as follows: (A) partial spraying, (B) full spraying, (C) initial wiping, and (D) wiping completion.

For endpoint evaluation, the four wiping conditions—full coverage, zigzag wiping, partial wiping, and uneven wiping—were repeated six times each. The proportions of sprayed, wiped, combined-treatment, and untreated areas were computed based on pixel segmentation.

2.4. Statistical analysis

Statistical analyses were performed using Prism 9 (GraphPad Inc., La Jolla, CA, USA) and the R software. Data are presented as the median and interquartile range (IQR). Each statistical test is detailed in the corresponding figure legend. Statistical significance was defined as p < 0.05.

3. Results

3.1. Integrated wipe-cleaning assessment system: system construction and operational characteristics

The integrated wipe-cleaning assessment system was successfully constructed and operated, as shown (Fig. 1A, B, and 1C). The RGB and thermal cameras captured the wiping and spraying activities without interfering with the operator movement or cabinet airflow. The schematic diagram (Fig. 1A) illustrates the noninvasive placement of the sensing modules.

The functional workflow of the system is summarized in the block diagram (Fig. 1B). During operation, the temperature drops detected by the thermal camera were clearly visualized as ETH-sprayed regions, and RGB imaging reliably captured the wiping motion across the work surface. The pressure sensor produced distinct electrical resistance signals in response to physical wiping.

The pressure-sensing module demonstrated clear discriminability between noncontact and contact conditions (Fig. 1C). The electrical resistance decreased sharply when the wiping implement applied pressure, surpassing the predefined threshold used for pressure detection. This allowed the system to reproducibly distinguish active wiping from light or accidental contact. To facilitate interpretation of the pressure-sensor signal, we performed a load-calibration experiment using a weighing scale. The averaged sensor output showed a consistent relationship with the applied load, facilitating the conversion of the signal into an approximate equivalent load in kilograms (Fig. 1C, lower right). This calibration provided a practical reference for interpreting the predefined pressure threshold used in the biological validation experiments. Together, these results confirm that all sensing modules operated as intended and that the system successfully integrated the information necessary for wipe-cleaning assessment.

3.2. Integrated wipe-cleaning assessment system: biological disinfection efficacy

Bacillus subtilis was used in a biological assay—to verify the disinfection performance of the proposed system. Three treatment conditions were compared: ETH spraying alone, wiping with no ETH, and combined-treatment with ETH spraying and pressure-based wiping (Fig. 2A). Agar plate cultures showed visible bacterial colonies in both ETH-only and wiping-only groups, whereas the combined-treatment group exhibited markedly reduced colony formation, including cases in which no detectable colonies were observed (Fig. 2B). Quantitative analysis demonstrated a significant reduction in CFUs in the combined-treatment group compared to the other two conditions (Fig. 2C).

3.3. Timeline-based and endpoint evaluation of the integrated wipe cleaning assessment system

Integrated wipe-cleaning assessment-system capability to track disinfection progression was evaluated using timeline-based and endpoint analysis. In timeline-based analysis (Fig. 3), the system successfully captured the sequential stages of the cleaning process, from partial ETH spraying to full coverage, and from initial wiping to the completion of zigzag wiping. Integrated data, including RGB, thermal, and pressure-sensor outputs, were synchronized to visually illustrate the classification of surface areas into sprayed, wiped, combination-treated, and untreated regions at each time point (Fig. 3).

Fig. 3.

Timeline-based process diagram in the zigzag wiping. (A–D) Sequential frames showing the wipe cleaning process during zigzag wiping, captured in a biosafety cabinet. (A) Partial ethanol spraying: only part of the surface is sprayed with disinfectant. (B) Full ethanol spraying: the entire surface has been sprayed. (C) Partial wiping: wiping is initiated but incomplete. (D) Zigzag wiping: the surface is wiped thoroughly using a zigzag motion. (A–D) Each panel displays the following outputs: Top left: RGB-camera image; Top right: orthographic projection; Middle left: thermal image; Middle right: thermal-difference image; Bottom middle: pressure-sensor signal; Bottom: output of the integrated assessment system, where green indicates sprayed areas, yellow indicates combined-treatment areas (spraying + pressure-based wiping), and red indicates untreated areas.

Four wiping conditions (full coverage, zigzag, partial, and uneven) were tested for endpoint-reproducibility evaluation (Fig. 4). Each condition was repeated independently, six times. The system generated classified output images (Fig. 4A), and the coverage area for each cleaning category was quantified as a percentage of the total surface area (Fig. 4B). In the combined-treatment areas, statistical analysis revealed significant differences in proportion between the full-coverage condition and all other conditions, demonstrating the sensitivity of the system to variations in cleaning performance.

Fig. 4.

Quantitative reproducibility analysis: reliability assessment under variable wiping conditions.

(A) Representative outputs of four wiping patterns—full coverage, zigzag, partial, and uneven—demonstrate ability of the system to differentiate between regions based on wipe-cleaning performance. In the output images, green represents sprayed areas, blue represents wiped areas without ethanol spraying, yellow indicates combined-treatment areas (spraying with pressure-based wiping), and red indicates untreated areas. (B) Quantitative analysis of each treatment was performed by calculating the proportion (%) of sprayed area, wiped area, and combined-treatment area based on pixel classification in the output images. Asterisks (*) indicate statistically significant differences (p < 0.05) compared to the full-coverage condition, as determined by Kruskal–Wallis test followed by Dunn's multiple comparisons test with Holm correction.

3.4. Correspondence between system-classified treatment regions and residual microbial contamination

To directly examine whether the regions classified by the system reflected the actual hygienic status, we performed a spatially resolved biological validation experiment (Fig. 5). After inoculation of the SUS surface with Bacillus subtilis (1.0 × 10³ CFU) and air-drying, ethanol was partially sprayed, and wiping was applied such that four distinct regions were generated on the same surface: sprayed-only, wiped-only, combined-treatment, and untreated areas (Fig. 5A). These regions were identified based on the integrated classification output and individually sampled for microbiological recovery.

Fig. 5.

Correspondence between system-classified treatment regions and residual microbial contamination.

(A) Experimental overview of the spatially resolved biological validation. Bacillus subtilis was inoculated onto a stainless steel (SUS) surface and air-dried. Ethanol was then partially sprayed, and wiping was performed to generate four distinct regions on the same surface: sprayed-only, wiped-only, combined-treatment, and untreated areas. The left panel shows the representative surface image prior to treatment, and the right panel shows the corresponding classification output generated by the integrated assessment system. Green indicates sprayed-only areas, blue indicates wiped-only areas, yellow indicates combined-treatment areas, and red indicates untreated areas. (B) Representative agar plates showing residual B. subtilis recovered from each classified region. (C) Quantitative comparison of residual microbial contamination among the four classified regions. The combined-treatment area showed the greatest reduction in residual bacteria. Asterisks (*) indicate statistically significant differences between groups (p < 0.05), as determined using the Kruskal–Wallis test followed by Dunn's multiple comparisons test with Holm correction. ND indicates “not detected,” and the dashed line indicates the detection limit.

Representative agar plates showed clear differences in residual microbial contamination among the four system-classified regions (Fig. 5B). Residual colonies remained detectable in the sprayed-only, wiped-only, and untreated areas, whereas the combined-treatment area showed the highest reduction in viable bacteria, including two samples in which no colonies were detected. Quantitative analysis confirmed that the combined-treatment area exhibited significantly lower residual contamination than the other classified regions (Fig. 5C). These findings demonstrate that the spatial classifications generated by the system correspond to biologically meaningful differences in residual contamination and support the validity of the integrated output as an indicator of local cleaning adequacy.

4. Discussion and conclusion

This study developed an integrated wipe-cleaning assessment system comprising an externally mounted RGB camera, thermal, and pressure-sensing modules, enabling nonintrusive and real-time monitoring of wiping operations in BSCs. The modular architecture allowed the simultaneous acquisition of visual, thermal, and mechanical signals, which were processed through a unified decision framework to classify surface conditions during cleaning. The pressure-sensing component could clearly distinguish between the contact and noncontact states, demonstrating its suitability for detecting meaningful wiping pressures. This framework addresses the growing need for quantifiable, recorded validation of cleaning processes, particularly in high-stake environments such as cell-processing facilities.

Biological validation using Bacillus subtilis demonstrated that ETH spraying alone or wiping without a disinfectant were insufficient to achieve complete bacterial removal, whereas a treatment that combined ETH spraying with pressure wiping resulted in a significant reduction in viable colonies (including cases with no detectable growth). These findings are consistent with those of previous reports emphasizing the importance of both mechanical and chemical actions during surface disinfection [7,14,15]. Under dry conditions, the contact area between the wiper and surface increases, resulting in a decrease in the applied pressure per unit area, which consequently reduces the cleaning effectiveness [7]. Therefore, wiping only achieves limited microbial reduction, whereas wet wiping markedly enhances biocidal efficiency by improving surface wetness and frictional contact [7,14]. In addition, the combination of spraying followed by wiping has been shown to be the most effective approach for sporicidal activity on SUS surfaces compared with spraying or wiping alone [15]. Taken together, these results indicate that mechanical wiping not only removes microbial contaminants but also amplifies the activity of the disinfectant through enhanced contact and distribution. The strong correlation observed in our system between the sensor-detected wiping pressure and biological decontamination reinforces the necessity of quantifying mechanical engagement for reliable surface disinfection in BSCs.

The integrated wipe-cleaning assessment system demonstrated reliable performance under a variety of wiping conditions, including full coverage and zigzag, partial, and uneven patterns. Timeline-based analysis confirmed that integrated sensing can capture the dynamic progression of the disinfection process, allowing visualization of the sprayed, wiped, and untreated areas in real time. This approach aligns with recent advances in sensor-based cleaning validation, where integrated imaging and pressure detection have been utilized to quantify the cleaning performance and ensure process traceability [16,17]. In particular, the use of thermal and optical imaging to detect cleaning coverage has been shown to improve reproducibility and reduce subjective bias in environmental disinfection [18]. Nevertheless, studies focusing specifically on pressure-based monitoring during manual wiping remain scarce, as most previous studies have relied on indirect indicators, such as visual inspection or chemical residue analysis. In robotic cleaning, pressure and tactile sensors have been utilized to optimize contact force and coverage uniformity, contributing to more consistent cleaning outcomes [19]. These developments support the feasibility of using pressure feedback as an effective indicator of cleaning completeness, although its application to manual disinfection remains limited. In addition to these technical advantages, visualizing the spray coverage and wiping trajectories may enhance the psychological assurance of operators during BSC work. Concerns regarding the completeness of cleaning operations have been documented in questionnaire-based assessments of routine disinfecting tasks [10,11], suggesting that personnel often perceive uncertainty regarding whether their manual actions sufficiently cover all target areas. Consistent with these observations, a prior work demonstrated substantial variability in both spray volume and spatial distribution when eight operators were instructed to apply disinfectant [7]. Unintentional inconsistencies, including peripheral underspraying or uneven initial deposition, represent latent sources of operator anxiety and potential contamination risk. By making these deviations visible in real-time, the integrated sensing system can reduce subjective uncertainty, support more confident task initiation, and promote a more consistent and deliberate disinfection mindset among operators. Despite the variability among repeated wiping operations and differences in operator techniques, the classification accuracy for combined-treatment areas in our system remained consistent across trials. Collectively, these findings suggest that integrating visual, thermal, and pressure feedback provides a reproducible and reliable framework for quantitative disinfection assessments in BSC operations.

The spatially resolved biological validation experiment further strengthened the practical relevance of the proposed system by showing that the system-classified regions were associated with different levels of residual microbial contamination. The combined-treatment area exhibited a significant reduction in viable bacteria, whereas the sprayed-only, wiped-only, and untreated areas retained detectable numbers of colonies. Although the combined-treatment area showed the most significant reduction effect, a few residual colonies were detected during some tests. This may be attributed to several factors, including the pressure threshold adopted in this study, which was defined as a practical prototype setting and may not yet represent the biologically optimal level required for complete microbial removal. Another factor is the deliberately high microbial load used to improve detection sensitivity. In the experiment, the inoculum was approximately 1000-fold higher than the level expected under actual contamination conditions in manufacturing environments, whereas microorganisms present in Grade A environments would generally be expected at considerably lower levels. Therefore, the occasional residual detection observed under these experimental conditions should be interpreted from the perspectives of both the stringent test setting and the currently empirical pressure-threshold setting. Accordingly, the system may be beneficial not only for visualizing cleaning operations in real time, but also for identifying insufficiently treated areas in a biologically meaningful manner.

This study had several limitations. First, the experiments were conducted under controlled laboratory conditions using Bacillus subtilis as a biological indicator and therefore may not fully represent real-world contamination complexity involving mixed microbial populations and surface residues containing abundant human-derived proteins and lipids. Secondly, although the system successfully quantified the pressure applied during wiping, the threshold for defining “sufficient pressure” lacks a biological basis. It remains unclear which specific pressure level ensures complete microbial removal, as this may vary depending on the microscopic surface conditions, such as the presence of scratches or unevenness, which can influence the retention of contaminants. Establishing biologically meaningful thresholds is an important step toward translating this system into a standardized validation tool. Additionally, the stick-type device was adopted as a simple and cost-effective approach for proof-of-concept development. Although this configuration enabled rapid prototyping, it may result in nonuniform pressure distribution across the sensor array. Future studies should investigate alternative device configurations, such as pad-type and flat-iron-like designs, to achieve more consistent and standardized pressure delivery during wiping.

In conclusion this study, we developed a sensing system that integrates RGB imaging, thermal sensing, and pressure detection to evaluate surface disinfection in BSCs. The system demonstrated both biological validity and operational reproducibility and successfully linked sensor signals with actual decontamination outcomes. By enabling a real-time and recordable assessment of wiping operations, it provides an objective and quantitative framework for verifying the cleaning performance.

In future applications, the proposed system can not only ensure and evaluate reproducible wiping performance but also serve as a real-time tracking and feedback platform. By visualizing the wiping trajectories and pressure distribution—information that is typically invisible to the naked eye—the system can facilitate onsite operational adjustments and post-process reviews. Moreover, such real-time visualization has the potential to enhance operator training and procedural standardization, similar to motion analysis and biofeedback systems that are widely utilized in motor learning and rehabilitation research [20,21]. The integration of these capabilities may further improve consistency, reduce operator-dependent variability, and contribute to skill transfer in BSC operations.

Collectively, this integrated sensing approach bridges the gap between manual cleaning and measurable validation, offering a practical direction for standardized, reproducible, and data-driven hygiene management in biosafety and cell processing environments.

Author contributions

MM, YK, YS, and JM; Data acquisition, analysis, and interpretation: MM; Manuscript drafting: MM, YK, YS, JM, and IS; Manuscript revision to address important intellectual content. All the authors have read and approved the final version of the manuscript.

Funding

This research was funded by a joint research grant from the NTT Corporation, Japan.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Mitsuru Mizuno reports financial support was provided by NTT corporation.