Intelligent In situ Printing of Multimaterial Bioinks for First-Aid Wound Care Guided by Eye-In-Hand Robot Technology

Abstract

INSIGHT (INtelligent in situ printing Guided by Eye-in-Hand robot Technology), an innovative computer vision-enabled system that combines a depth camera with a 6-degree of freedom robot arm, empowering it to identify arbitrary areas at various angles through real time adjustments and to enable volumetric printing performed by dynamic image recognition based on color and contour differences is presented. Continuous targeting of multiple wounds at different locations is achieved. The optimized pneumatic valve synchronized with the INSIGHT can print multiple inks with diverse rheological properties to fabricate scaffolds and bandages with the capacity to treat various types of wounds. The design of dual printed modes, such as extrusion and spray, can significantly decrease printing time for large-scale wounds on an ex vivo porcine model. INSIGHT demonstrates its ability to treat diabetic wounds, using a microgel-based ink possessing an inherent porous microstructure to facilitate cell infiltration. In vivo verification highlights its adaptability to enable customized care for rapid emergency treatment of trauma patients.

1. Introduction

Current in situ bioprinting technologies allow the creation of 3D structural organizations of engineered tissues through the use of sophisticated deposition in custom designed patterns to fill injured areas.^[1] This technology enables the deposition of bioinks for layer-by-layer building of complex constructs.^[2] However, utilization of a single-planar printing approach is not effective to address extensive or severe wounds possessing different shapes or anatomical locations, which ultimately impedes the printing process in clinical applications. Therefore, numerous research efforts have resulted in the development of more flexible printing systems, including 6-degree of freedom (DOF) robotic arms.^[3] This strategy exhibits several advantages over conventional 3-DOF bioprinting systems, facilitating a higher degree of integration.^[4] Printing systems that utilize a robot arm have achieved a superior level of integration between engineered and host tissues because a robot arm can directly inject bioinks into the injured area with clinically relevant dimensions, thereby filling complex tissues in a substantially more personalized, controllable, and reproducible manner. The use of robot arm-assisted in situ bioprinting^[5] has led to significant advances in direct dispensing of soft and viscoelastic biomaterials into damaged tissues, including bone,^[3a,6] cartilage,^[6a,7] skeletal muscle,^[8] and skin wounds.^[2d,9]

Prior to conducting in situ printing procedures, data regarding the anatomical information is required in order to ascertain spatial configurations for printing path determination. Several studies have demonstrated the feasibility of in situ printing by utilizing predesigned printing codes created through computational programs to fill defects,^[7a,9a,10] or after scanning with microcom- puted tomography,^[6d] or a 3D scanner.^[2d] (Table S1, Supporting Information) Nevertheless, notable discrepancies between the current research on in situ bioprinting and its suitability for application remain. This is primarily because it is challenging to customize the printing process according to the complex geometries and location of each patient’s defect using only predesigned codes and a 3-DOF robot arm. Moreover, traditional processes require multiple manual calibrations per defect, which can be time-consuming.

Recently, adaptive 3D printing has been introduced through the integration of a computer vision system and printing.^[3b,11] Indeed, the compositions of either monocular or binocular vision systems allow them to easily reconstruct the actual work- piece geometry after scanning.^[3b,5] Although these closed-loop feedback approaches enable real-time adaptation to randomly designed targets, eye-to-hand systems have a limited field of view, requiring multiple coordinate correlations to obtain geometric information. This limitation can result in incomplete 3D modeling of targeted volumetric regions, ultimately reducing accuracy.

In this study, we introduce an INtelligent in situ printing Guided by Eye-in-Hand robot Technology (INSIGHT). INSIGHT incorporates a depth camera directly attached to an end effector of a robot arm with 6-DOF. This configuration enables the system to perform scans of arbitrary topographical regions that would be unattainable using a single-planar printing setup with 3-DOF robot arm. The proposed INSIGHT system distinguishes itself with several key benefits over both traditional in situ printing technology and sequential printing which is including scanning process and printing separately. The complexity of wound geometry is unpredictable, with potential crevices, obstacles, and over-hangs that a simpler robot (3-DOF)-based in situ printing cannot accommodate. Similarly, the sequential printing approach faces challenges when dealing with protruding parts at the wound, making it difficult to insert constructs without causing damage. Limitations arise when attempting to maintain 3D structures in large (> 5 or 10 cm) and thick-scale constructs due to the mechanical properties of bioinks. Moreover, transferring hydrogel-based printed constructs poses a risk of damage during the process, emphasizing the superior effectiveness of directly printing constructs over the sequential strategy.

INSIGHT is driven by artificial intelligence (AI) and functions through precise identification of the targeted regions, management of the robot arm’s motion, and simultaneous dispensation of accurate quantities of bioinks to the impacted site. In addition, the automatic generation of printing trajectories facilitated by computer vision algorithms expands the scope of the targets including not only simple 2D outlines but also complex 3D structures within damaged tissues. From a biomaterial perspective in the context of bioinks, it is imperative to select materials possessing multifaceted functionalities.^[12] These attributes are essential for addressing a wide range of injuries, each with its own distinct treatment goals. For instance, when dealing with traumatic wounds characterized by severe deep tissue damage or exposed muscles, the choice of ink with robustness, suitable for internal layering to effectively fill the wound, is crucial.^[8a] For chronic wounds, such as diabetic foot ulcers, a porous material-based ink can facilitate host tissue migration^[13] by possessing interconnective porous structure while simultaneously providing a bandage material within the printing ink to prevent secondary infections. This approach becomes particularly advantageous when dealing with large or complex-shaped wounds, as it allows the creation of custom-designed, one-step printed dressings. Therefore, in order to generalize the technology for providing first aid treatment to diverse wounds, it is demonstrated through INSIGHT that the synchronized pneumatic valve control enables the application of various inks tailored to specific purposes by adjustment of environmental factors such as pneumatic pressures. This showcases the excellence of our technological capabilities in each application area, including the printing application of natural polymer blend ink, micro-granular gel-based ink, and a biocompatible elastomeric ink. Furthermore, leveraging the degrees of freedom of the robot arm we employ an independent multi-nozzle system, which can print inks with heterogeneous properties without the occurrence of mixing, thereby preserving the functionality of each ink efficiently. The integration of dual printing modes – extrusion and spray printing – permits fast switching and printing of different bioinks, resulting in the fabrication of multi-material constructs at large-scale wounds that are unachievable using conventional printheads. Consequently, the IN-SIGHT system demonstrates enhanced effectiveness in rapidly covering wound areas, surpassing the efficiency of the sequential extrusion-based printing strategy by directly spraying the defect. Finally, we demonstrate the feasibility of INSIGHT for treating diabetic wounds, through the deposition of microgel-based bioinks, onto a db/db mouse model. This innovative approach overcomes the major hurdles associated with conventional in situ bioprinting, such as the time-consuming creation of printing paths and the dependence on pre-designed geometries, by enabling immediate and rapid printing.

2. Results

2.1. Overview of INSIGHT and its Scan Accuracy

Figure 1 provides a detailed overview of INSIGHT’s high level of autonomy and patient-specific functionality, coupled with substantial enhancements to its clinical capabilities. The INSIGHT system as illustrated in Figure 1A, consists of three key components: a visual guidance system that utilizes image recognition algorithms, a 6-DOF robotic arm control system, and a printhead equipped with computer-controlled pneumatic valves.

Figure 1.

Overall process of INtelligent in situ printing Guided by Eye-in-Hand robot Technology (INSIGHT) for first aid targeted wound treatments: A) Overall setup of INSIGHT components and process flow including (1) vision recognition, (2) computational geometry to (3) pneumatic valve controller for printing. B) Dynamic scanning process assisted by visual guidance through mounted depth camera and C) different bioprinting strategies including extrusion and spray method and D) various types of multimateral inks for different functions in wound repair.

INSIGHT is controlled via software written in the Python programming language, including a manufacturer-provided camera library and custom-developed libraries to position the robotic arm and toggle valve openings through commands via ether- net, as depicted in Figure S1 (Supporting Information). Scanning with the depth camera mounted on INSIGHT precisely recognizes spatial information of one or multiple targeted wounds on a trauma patient, especially their location, contours, colors, and depth in a continuous manner (Figure 1B). This capacity allows real-time and quick adaptations to accommodate changes in the geometries of the affected areas after printing. It also enables complete filling of deeply injured regions. After scanning, the desired bioinks are selected according to the size and depth of the wound recognition process performed by INSIGHT and extruded or sprayed at the wound area (Figure 1C).

These three subsystems of INSIGHT work seamlessly to precisely identify the affected area, control the printing process, and dispense required volumes of materials. The visual guidance system is aided by a depth camera, which is attached to the peripheral head of the robot arm and is composed of RGB color sensors and stereo imagers. It captures color images and stereoscopic depth imagery, which are then processed using functions in the OpenCV library to isolate the region of interest.^[3b] By orienting the camera toward the targeted object or defect, surface information, including color images and stereoscopic depth imagery, is acquired using a 3D recognition mechanism based on the principle of binocular disparity.^[14] This assembled system allows for dynamic collection of multiple diverse views of the targeted objects by rotating its arms in 6-DOF. This results in few spatial scanning restrictions for object positions. Notably, this coordinative assembly of dynamic recognition via visual guidance and the flexible motions of the 6-DOF arm synergistically enable larger angles for scanning and printing. Further, to enhance the versatility of INSIGHT to administer initial medical care to diverse wound scenarios, different types of printing modes using several biomaterial-based inks are demonstrated (Figure 1D). In particular, a blend of natural polymers, a micro-granular gel-based ink, and a biocompatible elastomeric ink, each possessing distinct rheological characteristics are employed to provide different functionalities such as creating an engineered volumetric construct in an aligned manner, inducing fast angiogenesis, or ensuring comprehensive coverage of the wound to prevent infections. In addition, adaptable printheads equipped with multiple reservoirs for these bioinks are incorporated onto INSIGHT. The initial step of INSIGHT is to identify the area of interest, assisted by the visual guidance system through the mounted depth camera (Figure 2A).

Figure 2.

Scanning properties of INSIGHT: A) Demonstration of dynamic recognition process of the artificial wound (scale bar = 4 cm) and B) demonstration of different path planning after recognizing the artificial wound. C) Difference of scan processing time based on scanned wound area (n = 5). D) Scanned example of skin wound, its pointclouds, and the schematic illustration of comparison of scanned values (area and depth of wounds) between the real values (scale bar = 2 cm). E) Area error ratio of different skin and wound color combinations (n = 10). F) Depth error ratio of different skin and wound color combinations (n = 10). G) Schematic illustration of tilted wounds at different angles. H) Area error ratio at different scan angles (n = 10). I) Depth error ratio at different scan angles (n = 10).

Subsequently, a mask is generated to detect the contours of the targeted area with color extraction. For instance, the skin wounds have distinctive red colors from the surrounding uninjured skin, so the contours of the wounds directly within the camera view can be precisely isolated by specific computational geometric processing. Then, pointclouds which lie within the isolated contours are implemented in the Open3D library, displaying the position information. Alpha-shapes are then utilized to generate a polygon that bounds the isolated object obtained from the mask, using the pointcloud points of the defect, in order to identify the clusters. The alpha parameter in the alpha shape Python library is employed to adjust the tightness of the resulting polygon, enabling it to closely resemble the outline of the defect, more accurately than other computational geometry approaches, such as a convex hull algorithm. Next, all points in the pointcloud, obtained from uniformly sampling the mesh of the original pointcloud, are iterated through to determine the points inclusivity with respect to the alpha shape. Points inclusive to the alpha shape are utilized to construct a robot travel path, whereas points exclusive to the uniformly sampled pointcloud are discarded. This is a computational geometry technique used to create a concave shape that best fits a set of points.^[15] This approach provides greater control over the number of points in the printing path and the robot’s trajectory. The density of path points was found to be positively associated with the size of the grid, and the line, and point space.^[6e] The line and grid values are modified to regulate the quantity of points to be printed within the reconstructed mesh.

Through flexible adjustments of the line and grid quantities for printing within the excluded target region’s mesh area, diverse printing outcomes can be achieved with various strategies. This adaptability in the number of lines and grids allows varying densities in the printed results, depending on the thickness of the dispensed ink. The subsequent images illustrate distinct pointcloud results that follow from the incremental increase in the number of lines and grids within the wound area, corresponding to a volume of 3 mm³, ranging from 50 to 200 (Figure S2, Supporting Information). Furthermore, by selectively organizing points from the pointclouds into regular arrays and aligning them in a specific axis direction, it is possible to establish the printing direction along either the x-axis or the y-axis (Figure 2B). This process enables efficient and accurate printing. We observed that the total scan processing time, from scanning an area as small as 1 cm² to as large as 100 cm², increased from 46.3 s to 172. 9 s (Figure 2C). As a result, even with the substantially larger 100 cm² area scan, the geometry processing step requires less than 3 min. During capturing the pointclouds, color extraction, and mesh reconstruction steps, there was no significant difference in processing times. regardless of the scanned wound size, whereas, when the concave hull algorithm was applied, processing time significantly increased due to increased size of the target (Figure S3, Supporting Information).

Additionally, we investigated the scan field of view in the x-y plane by varying the scan position in the z-direction. When the scan height (h) was increased from 20 to 55 cm in 5 cm intervals (Figure S4, Supporting Information), the scan field of view in the x-y plane also expanded from 0.05 m² to ≈0.60 m², effectively covering the entire table area. Regardless of the change to scan heights, the measurement of both length and width of the black squares in the targeted board was high accurate. The depth camera of INSIGHT exhibited limitations in detecting thicknesses within the 2–3 mm range at scan heights exceeding 30 cm. To overcome this limitation, a potential solution is to conduct scans at an elevation of 20 cm, where the camera can accurately capture thicknesses of 2–3 mm. By performing scans across multiple x-y plane regions and then merging, the resulting pointclouds, a broader range of information can be obtained without compromising the resolution of the scan depth.

To perform image processing, the images captured by the camera were converted into the Hue, Saturation, and Value (HSV) color spectrum. Then, the pixels from the images were isolated falling within the pre-defined lower and upper bounds, in order to create a preliminary mask. The outline contours passing through the mask were extracted to refine the mask, effectively compensating for potential issues caused by light reflections or other factors. Then the pixels in the depth image were analyzed using this refined mask. Points corresponding to black regions in the mask are removed from the pointclouds, while points within the white regions are retained. This procedure ensures the inclusion of only pertinent points in the final pointclouds, leading to improved precision and data quality. Extracted pointclouds with the color map indicating the depths in Figure 2D show an example of color isolation, based on this process through the canny edge detection algorithm. Taking into account diverse skin colors and wound types, we conducted scans for a total of 36 color combinations to assess the accuracy of identifying the target wound area (Figure 2E) and depth (Figure 2F). Notably, all color combinations demonstrated an accuracy exceeding 95%, achieving high precision and the potential to precisely target and perform in situ printing at the correct locations within specified tolerance levels at the unit-level (Figure S5, Supporting Information). Depths of the target were also recognized in a highly accurate manner. However, a slight deviation in depth recognition was observed when identifying the narrow depth (Figure S6, Supporting Information). In addition, regardless of differences in depth, overall processing time showed no difference.

Due to the camera’s field of view being equivalent to the robot’s range of motion, the assembled system facilitates dynamic capture of multiple diverse views of targeted objects by maneuvering its arms along 6 different axes. INSIGHT’s field of view is dependent on the combination of the robot arm’s angular range of motion and the camera’s capabilities, which specify a scan distance of 50 cm between the camera and target with a depth accuracy within 2%. Despite changes in scanning angles from −75 ° to 75 ° in intervals of 15 °, as presented in Figure 2G and Figure S7 (Supporting Information), the scanned area, after undergoing target region color isolation, exhibited a remarkable accuracy of 99% (Figure 2H). However, for the recognized depth, some deviations were observed, although it still demonstrated relatively high accuracy (Figure 2I). These deviations are attributed to uncertainties inherent in cameras, which can be influenced by environmental factors such as lighting conditions. Moreover, with the in situ printing ink possessing an average thickness of around 1.5–2 mm, an error of 5% or less would result in a height difference of ≈100 μm, which may be insignificant.

2.2. Dynamic Recognition and In Situ Printing

Since INSIGHT considers the coordinates of the scanned surface as the initial points to locate directly, it offers the notable benefit of removing the necessity of calibrating coordinates. The algorithm involves a coordinate translation process to accurately target the object’s position. Initially, pixel-based camera coordinates are transformed into effector coordinates of the robot arm. This conversion involves mapping x and y pixel values to their corresponding x and y coordinates, while the z data from the depth camera determines the camera’s height relative to the objects. To demonstrate INSIGHT’s positioning functionality, we utilized the depth camera to capture an image of a keyboard, followed by reconstruction of the keyboard’s mesh (Figure S8, Supporting Information). Then, eight consecutive points were carefully selected, and the robot arm precisely moved to each of these positions and performed clicking motions, by utilizing accurate coordinate transformation process (Movie S1, Supporting Information).

After positional information is translated into robot coordinates, a computer-controlled pneumatic valve, which connects the bioink reservoirs mounted on the opposite side of the end effector, synchronously opens and closes to follow an automatically generated printing pattern to quickly perform in situ printing. In this study, a bioink formulation consisting of alginate, gelatin, and cellulose nanofiber (CNF) was optimized (Figure 3A), since CNF itself exhibited the shear-thinning characteristic which is essential for the extrusion-based printing.^[16] After printing, the printed construct was crosslinked via ionic crosslinking by adding 3 mL of 60 mm CaCl₂ solution on top of the construct. Superior printability was achieved, indicating a notable shear-thinning property as shown in Figure 3B.

Figure 3.

Demonstration of dynamic recognition based on color extraction and contour detection, and in situ printing process: A) Schematic illustration of the ink composed of gelatin, alginate, and CNF. B) Viscosity of the ink as a function of shear rate. C) Viscosity recovery test under low and high strain cycles. D) Viscosity of the ink as function of temperature. E) Image processing flow including image recognition, mask generation, mesh reconstruction, and in situ printing of the pattern motif (scale bar = 1 cm). F) Automatically generated pointclouds for random examples of wounds, showing precise location recognition and G) continuous robot movement images at each example for in situ printing (scale bar = 1 cm, 2 cm, 4 cm, respectively). Demonstration of dynamic scanning and function recovery. H) Photos of comparative image recognition between (i) original image and (ii) the puzzles missing two pieces (scale bar = 2 cm). I) Isolated contours (inset) and reconstructed mesh of the missing part. J) Photo of in situ printing of isolated missing puzzle parts (scale bar = 2 cm). K) Photo of recovered function confirmation after in situ printing of missing puzzle parts (scale bar = 2 cm).

From Figure 3C, this formulation of inks exhibited noticeable viscosity recovery after destructive extrusion-like deformation. Moreover, the transition between these states was maintained over multiple cycles. Thus, they could be employed for filling deep wounds by stacking of multi-layer structures. An excessive concentration of gelatin can have adverse effects on the printing process, potentially causing issues like nozzle blockage or uneven flow of the bioink.^[16b] Thus, the ink containing 2% gelatin was selected for further in depth analysis to minimize the clogging issue, within the optimized low-viscosity ink according to the literature.^[17] We then characterized the rheological properties of the CNF-laden ink. Indeed, a comprehensive assessment considering the entire viscosity spectrum and temperature sweep test (Figure 3D) revealed that 2% composition of gelatin is less affected by temperature.

An example of a pattern motif template was captured by IN-SIGHT, and the algorithm automatically generated its mesh by applying the concave hull algorithm to extract only the red color of the motif (Figure 3E). The coordinate transformation process measures the distance between the center of the depth camera and nozzle, attached to the robot end effector (Figure S9A, Supporting Information). To print accurately, we accounted for the ink thickness (≈1–2 mm) at ≈10 pounds per square inchi (psi) between the needle tip and the printing surface. To assess the repeatability of dynamic position recognition, we compared the actual origin of the points with the robot’s pointing positions (Figure S9B, Supporting Information). Using a continuous recognition and printing approach for wounds with diverse sizes and depths, we tested the system’s dynamic ability to recognize locations while simultaneously gathering information about the wound’s depth and area (Figure 3F).

By employing precise pneumatic valve control, the system enables accurate ink deposition within the target wounds, independent of their position, demonstrating its capacity for precise in situ printing (Figure 3G; Movie S2, Supporting Information). To achieve dense printing after recognition of an example of motif, the interval between points in the mesh was experimentally set to match the thickness of the inks extruded from the needle, with a size of 20 G (Figure S10, Supporting Information). The printing direction can be easily accommodated either in the x or y direction, precisely following the mesh patterns that we obtained through image recognition (Movies S3 and S4, Supporting In- formation). These results provide strong evidence of the printing pattern, confirming that the robot arm’s movements closely follow the automatically generated printing path determined by our visual guidance system, which dynamically recognizes the targeted objects.

2.3. Demonstration of Dynamic Recognition with Contour Comparison and Function Recovery After In Situ Printing

To achieve precise in situ bioprinting that can target damaged areas regardless of external factors such as placement and dimensions, it is crucial to dynamically recognize the missing part and gather information about its location and contours. The ultimate objective of in situ bioprinting is to restore structural integrity with anatomical continuity and recover tissue function. To this end, we have herein devised a sophisticated algorithm that utilizes INSIGHT to identify the missing part by comparing an object with its complete original geometry to one that has lost a section. To demonstrate the systematic recovery process and dynamic recognition, we used a quick response (QR) puzzle as an example for scanning (Figure 3H). As shown in Figure 3I, the contours of the two missing adjacent puzzle pieces were automatically generated by INSIGHT by identifying differences in color and contours between the arrays of the original puzzle and the one with two pieces removed. The algorithm accurately recognized the black squares of the QR pattern in the missing pieces, by image recognition along with comparison, which were ≈5 mm × 5 mm in size, in accurate positions. Then, precise high-resolution in situ printing was con- ducted in a rapid manner (Figure 3J), ultimately recovering the QR code (Figure 3K; Movie S5, Supporting Information). IN-SIGHT demonstrated its ability to successfully recognize and restore the function even in instances with two sections located far apart (Figure S11 and Movie S6, Supporting Information).

2.4. Scanning and Printing by INSIGHT at Different Angles

In real wound scenarios, the color and complexity of the wound varies, with large or deep wounds involving not just skin but also muscles and bones. Each body part possesses distinct colors, making complex wound shape recognition dependent on the camera position. Therefore, a series of ex vivo experiments was carried out to validate the processing ability of our system. In ex vivo rodent models, we verified the scanning and positioning resolution by creating wounds of different sizes (Figure S12 and Movie S7, Supporting Information). Due to tissue elasticity and gravity, the wound shapes differed from the punch shapes, and their depths varied as well. However, INSIGHT accurately scanned and printed the wounds in their exact positions, illustrating the potential of our system for real application and in situ bioprinting for small wounds. The pointclouds in Figure 4A demon- strate the system’s ability to fill large and severely curved defect regions, from general top view scans. In addition, we successfully generated precise in situ printing processes in the wounds located at the side of the subject that cannot be viewed from above with a 3 DOF printer by tilting at +/−30 °. For instance, in our scanning process, we assumed a tilted angle to capture images of the wound from different angles (−30 ° and 0 °) specifically focusing on the lateral regions of the target. After combining these images, we created a merged pointcloud that represents the wound in that particular region. To identify clusters within this pointcloud, we utilized the density-based spatial clustering of applications with noise (DBSCAN) algorithm in Python. To define the boundary surrounding the isolated wounds, we employed the alpha shapes method with the alpha parameter from the alpha shape Python library. This parameter allowed us to adjust the tightness of the resulting polygon, ensuring that it closely matched the wound’s outline with higher precision compared to other computational geometry approaches, including the convex hull algorithm (Movie S8, Supporting Information).

Figure 4.

In situ printing in wounds at different location at tilted motions: A) Printing process at (i) −30 ° (scale bar = 5 cm, 1 cm, 1 cm in clockwise), and the scanned pointcloud (ii) 0 ° (scale bar = 5 cm, 5 cm, 1 cm in clockwise), and the scanned pointcloud, and (iii) +30 ° in × direction (scale bar = 5 cm, 2 cm, 1 cm in clockwise) and the scanned pointcloud. B) Comparison of single- and multi-angle point cloud detection of wounds possessing complex positions and shapes and print results. The blue arrows indicated the missed part in the extracted point cloud and printed results.

Given the inherent complexity and non-planar nature of wound geometries, it is important to emphasize the system’s performance using wounds that feature crevices, obstacles, and overhangs. Another example exhibiting this requirement would be a wound with notable width and depth ranging from the top of the arm to the side, where a roll-type rotation of the printhead would be required to access all parts of the wound. The 3-DOF would be sufficient within a limited, reasonable range of orientations, however any object located outside this boundary or positioned in a way that hinders proper perception due to the camera’s angle and placement cannot be fully recorded by the camera sensor, resulting in data loss. Our INSIGHT aims to provide solutions that go beyond these limitations. It can be enhanced by gathering information by repositioning and reorienting the camera around the target. This approach allows for multiple perspectives, thereby rendering previously inaccessible or obstructed areas visible. To illustrate the contrasting outcomes of employing the pointcloud combination process as opposed to using a single pointcloud capture, two tests were conducted as depicted in Figure 4B.

In the case of the pointcloud generated from a single-angle capture, it was unable to reach certain areas that remained obstructed from the camera’s view. Consequently, this led to gaps in the pointcloud data, lacking in depth information. Since the printing path was not designed to account for these unrecorded areas, the subsequent printing process was negatively impacted, resulting in empty spaces where the ink could not reach. On the other hand, the pointcloud combination method successfully addressed the issue. It compensated for the missing information in first capture at 0° by integrating the pointcloud data from the 30° and 40° captures, which managed to capture the previously inaccessible points. This integration resulted in a comprehensive printing path that covered the entire targeted area with a detailed representation of the wound. This improvement was particularly valuable for wounds with intricate shapes that were challenging to recognize using single-angle pointcloud detection. As a result, we could generate precise printing paths that closely followed the complex geometry of the wounds, significantly influencing the quality of the printing outcomes. The 6-DOF feature of INSIGHT allowed capturing images of wounds from multiple angles, enhancing the accuracy of imaging.

2.5. Dynamic Depth Recognition and In Situ Multi Layer Printing

Designing and executing motion paths that ensure stable contact between robots and the complex surfaces of the human body remains a critical challenge in current research. This challenge is further exacerbated by the significant anatomical variations found in the human population, including uneven skin geometries. Thus, we have intensively examined robotic path planning by scanning with INSIGHT the wound-like phantoms placed on substrates with diverse curvatures. We demonstrated multiple printing iterations with rescanning after each layer was printed. Utilizing the z data obtained from the depth camera allows for the automatic adjustment of the camera’s height relative to the objects. This adjustment accounts for the increased height, corresponding to the increase in printed ink thickness. The system achieves this by rescanning the printed layer after each printing iteration, facilitating the seamless execution of consecutive multi-layer printing. To clearly visualize the printing direction, following the wound curvature, the numbers of the line and grid within the wound pointclouds were adjusted as depicted in Figure S13 (Supporting Information). By carefully following the wound’s surface, the ink is dispensed and fills the cavity effectively. This printing strategy proves useful when dealing with the same tissue in multiple layers (Figure S14, Supporting Information). After repeating the printing process four times, the cavity is adequately filled with ink, resulting in minimal height differences when compared to the uninjured skin (Figure 5A; Movie S9, Supporting Information). The total volume printed within the wound was measured until the overall ink height precisely matched that of the surrounding skin tissue, which was found to be highly consistent and reproducible with the actual empty volume present in the wound (≈3 mL) (Figures 5B,C; S15A, Supporting Information).

Figure 5.

Multi-layer in situ printing process by dynamic recognition of INSIGHT: A) Top view images and scanned pointclouds at each printing step for in situ printing within the wound area (scale bar: 4 cm). B) Photos of cross-sectioned and top-viewed printed construct after crosslinking by CaCl₂ (scale bar: 1 cm) and C) weight of the inks (n = 3). D) Schematic illustration of the nozzle with robot movement in skipping mode based on distance between each printing point. E) Side view of photos and top view of scanned pointclouds, and top view of printed results after dynamic recognition of INSIGHT for three different wounds on non-planar surfaces (scale bar: 2 cm).

Since it is known that skin exhibits viscoelastic characteristics when subjected to mild shear forces within a spectrum of physiologically significant frequencies, we examined the rheological response of the ink after undergoing crosslinking, as illustrated in Figure S15B,C, Supporting Information). The storage modulus value closely resembled that of the dermal layer around 440 –6620 Pa,^[18] suggesting the potential to attain a comparable elasticity, thereby minimizing the risk of damage of printed constructs caused by tissue compression or friction during daily activities. Various designs of the wound curvatures were prepared using several substrates, which were printed using polylactic acid (PLA), as depicted in Figure S16 (Supporting Information). The flexible wound phantoms placed on top of the various non-planar substrates exhibited different depth distribution within the wounds. An alternative bottom-up approach was chosen, and each print step was sectioned at intervals of 2 mm in height, taking into account the average ink thickness. Poor print quality can be explained by the occurrence of excess ink being dragged across the injured region, potentially leading to additional injury caused by the nozzle coming into contact with the wound. To address these issues and prevent the possibility of secondary injury, we implemented a solution for the robot movement during printing. Before printing a filament, the robot hovers at a safe distance of ≈2 cm above the wound. This hovering action is done regularly as a safety measure (Figure 5D). Additionally, the system autonomously recognizes the distance between the points that need to be printed. After this recognition step, a slight upward coordinate translation is executed for the nozzle, ensuring it remains at an acceptable distance from the wound surface. Finally, complex curved wounds were successfully filled by consecutive dynamic recognition (Figure 5E; Movie S10, Supporting Information).

2.6. Dual Printing Mode Compatibility demonstration Using ex vivo Model and Characterization of Spray Printing

In order to attain advanced functionality before developing into a clinical application, which swiftly scans and automatically creates the printing routes in a matter of seconds, it is crucial to solve the issue of expanding target size. This is considered a bottleneck, leading to extended duration for extrusion-based printing. As depicted in Figure 6A, a porcine wound model was chosen to demonstrate the different wound types in terms of the depth and the shape. For instance, through dynamic recognition of deep wounds of ≈10 cm thickness, as presented in Figure 6B and Movie S11 (Supporting Information), the extrusion-based printing mode was chosen to fill the defect effectively. Successful completion of the printing task in 10 iterations was achieved by employing iterative and dynamic scanning techniques and the automatic generation of the printing path. In contrast, when confronted with large wound areas that expanded significantly in the x-y plane and were shallow, the spraying mode is selected (Figure 6B).

Figure 6.

Dual printing mode of INSIGHT using ex vivo model and characterization of spray printing: A) Ex vivo setup using a porcine model in the operating room. Photos of in situ printing process and the results of printing for both B) extrusion and C) spraying using ex vivo porcine wound model (scale bar 1 cm). D) Printing time difference between two modes at different top-scanned wound area. E) An example of spray printing on top of the burn wounds using ex vivo model. (i) photos of the large burn wound, (ii) Numbers of line and grid adjustment for spraying and (iii) In situ spray printing results (scale bar 5 cm). F) Spraying particle size at different location, based on distance from the center. G) Schematic illustration of spraying for burn wound application. H) Spray simulated results at height of 5 cm. I) Simulated results of particle distribution and size at height of 5 cm. J) Simulated results of pressure distribution when printed in spraying mode on top of the wound at height of 5 cm.

The selection of different printing modes required incorporating additional air pressing steps during ink extrusion to ensure proper breakdown of the bioinks. Notably, our unique rheological bioinks consisting of gelatin, alginate, and CNF, enabled the effective deposition of relatively thick layers using the spraying mode. By integrating this versatile dual-mode printhead for spray printing mode, we achieved rapid and comprehensive coverage (Figure 6C), even for larger targets, exemplified by the 10 cm × 10 cm scenario in such a rapid manner (12 times faster than extrusion printing) as shown in Figure 6D. The spraying printing method employed here proves to be an efficient strategy for treating burn wounds, as illustrated in Figure 6G.

To achieve this performance, we reduced the number of lines and grid points during the geometric process, thereby increasing the spacing between each point. This adjustment was necessary to accommodate the spatial distribution of ink during spray printing (Movie S12, Supporting Information), which ultimately affects the accuracy and uniformity of ink coverage following the wound edge (Figure 6E). Taking into account the value of viscosity and the flow rate during spraying, a simulation was conducted to verify the particle distribution and the size of the sprayed inks, at one time point during spraying, when the height of the printing was 5 cm (Figure 6I). Based on the air blast atomizer model (Figure S17, Supporting Information), depicted in Figure 6H and Figure 6I, the sprayed ink particles ranged around 56 μm and were distributed within an 18.81 cm² area. In this model, we do not account for particle aggregation, whereas in experiments, particles have been observed to aggregate, appearing larger (Figure 6F). The particle sizes exhibited a spectrum of values as a function of distance from the center, extending from 0.5 cm to 2 cm, with sizes varying from 903 μm to 283 μm. Another plausible explanation is the omission of viscosity considerations, which may influence breakup dynamics, leading to significantly smaller sizes compared to the particle sizes observed in experimental results, as illustrated in Figure S17J,K (Supporting Information). The simulation indicated that the maximum pressure at the center of the wound area after spraying was 6.75×10⁻⁵ Pa (Figure 6J), which is significantly below the levels associated with tactile sensitivity or the skin pain pressure threshold,^[19] eliminating possible pain during the ink deposition.

2.7. Application of Bandaid Printing

An additional functionality of INSIGHT is the ability to resize the mask area during the printing of the final layer to effectively cover the whole wound region (Figure 7A). In this case, a new structure of an air-powered system with 6-DOF was assembled, where the ink that needs to be printed at that stage is placed separately (Figure 7B), to avoid the mixing of distinct viscoelastic properties of inks. For printing the band-aid, poly(glycerol-co-sebacate) acrylate (PGSA) polymer was used (Figure 7C). Photocurable PGSA was previously introduced to show its potential of 3D printing to create biocompatible and elastomeric tissue substitutes.^[20] Herein, we first demonstrated dynamically printed bandages that can completely encase the wounds using multimaterial inks. While the rheological behavior of the PGSA ink displayed a purely Newtonian response, it was still feasible to ensure the printability of the ink for wound coverage applications (Figure 7D).

Figure 7.

Multi-layer in situ printing process by Dynamic Recognition of INSIGHT for Bandaid printing: A) Schematic illustration of multi-layer printed structure, showing the size of the bandaid printing area has been enlarged to cover the wound area after 2nd layer in situ printing. B) Schematic illustration of syringe holder design for multi-inks printing. C) Schematic illustration of wound covering ink based on PGSA. D) Viscosity of PGSA based inks as a function of shear rate. E) Photos of each layer printing (scale bar 4 cm), by changing the angles of the syringe holders as utilizing the degree of freedom of the robot arm and captured images after printing at each step (scale bar 2 cm) with pointclouds. F) Photos after bandaid printing (scale bar: 2 cm).

This was evident in the printed outcomes, which closely resembled the size of the band-aid mask, indicating that the ink did not exhibit significant lateral spreading after printing. After printing PGSA ink as pre-polymer state on top of wound scaffolds filled with 1st and 2nd gelatin based bioinks, the ink was photo-crosslinked for 30 s at 800 mW cm⁻² of UV light. To enable the acrylate-functionalized inks to be tissue-adhesive via covalent bonding as a result of the radicals produced during the curing process,^[21] the printing area for forming the band-aid was automatically resized by increasing the mask size by 1.5 times compared to the original size of the mask (Figure S18, Supporting Information).

The command to rotate the end effector of the robot at a 45° angle in both directions, individually synchronized with three separate valves, was implemented as a proof of concept to enable precise control and simultaneous manipulation of multiple inks during these three printing steps (Figure 7E). Three separated valves were individually synchronized with three syringes. Based on the viscosity of those multi-inks and printability of the inks, different levels of air pressure were applied (10 psi for wound filling, and 5 psi for band-aid printing). After crosslinking, the absence of undesired ink flows was verified, as they were effectively covered by the bandaid printed films (Figure 7F). This confirmed the successful completion of the multi-layer printing process.

2.8. In Vivo Diabetic Wound Healing Efficacy Using INSIGHT

The wound healing efficacy of INSIGHT was validated using an established in vivo model of impaired diabetic wound healing, the db/db mouse. Biocompatible bioinks with microstructures facilitating the influx of cells and cytokines, chemokines and growth factors responsible for wound healing were utilized. Briefly, the microgels composed of gelatin, methacrylated gelatin (GelMA), and polyethylene oxide (PEO) were prepared through two phase aqueous emulsion systems in order to possess good biocompatibility (Figure 8A).^[12c,13,22]

Figure 8.

In situ printing on in vivo diabetic wound model, using the microgel inks. A) Photos of in situ printing process after recognition of the wound using INSIGHT. (i) photo of the setup, (ii) pointclouds of the scanned wound, and iii) photo of in-situ printing (scale bar 1 cm), and schematic illustration of the microgel ink structure composed of gelatin and GelMA. B) Confocal image of microgel inks (scale bar: 10 μm). C) Live/dead image of HDFs on microgel samples after 3 days of culturing with the density of 0.2 m mL⁻¹ (scale bar 200 μm). D) Viscosity graph of the microgel ink as a function of a shear rate. E) Modulus of the microgel ink in a strain sweep mode. F) Wound closure images at different time points (scale bar 5 mm, and microgel inks deposited on the wound were marked with red arrows). G) Wound closure area graph at different time points. Masson’s Trichrome Stained images of H) TD only group and I) microgel ink printed group at day 10 (scale bar 1 mm and scale bar 200 μm). J) Area of hyperproliferative epidermis for TD only and Microgel ink group (* p < 0.05).

The interconnected microporosity within the micro-sized gel granules with the size from 23 to 53 μm facilitated the movement of cells at the wound site, leading to enhanced incorporation with the adjacent tissue.^[23] In our study, spherical shaped microgels with an average size of 3 μm yielded excellent printability (Figure 8B). Gelatin’s distinctive critical temperature point enables particle generation at ambient room temperature, and these non-crosslinked sacrificial particles can be eliminated when exposed to body temperature, resulting in the formation of a porous network. This exceptional pore forming characteristic of the emulsion-based bioink promoted human dermal fibroblasts (HDFs) proliferation and spreading on the surface of the crosslinked construct, with good elasticity as depicted in Figure 8C. In situ printing was conducted after wounding of anesthetized mice, with the printing performed ≈1 mm higher than the recognized height to prevent nozzle contact with the wound during breathing-induced movements.

Once in situ printing was complete, UV light was employed to crosslink the bioinks. The entire process, from scanning to printing path generation, took ≈30 s per wound, and the printing itself lasted ≈30 s, allowing for rapid ink deposition due to its good shear thinning behavior (Figure 8D) to maintain its elasticity under low strain, until ≈7 Pa of shear stress, highlighting the microgel’s internal resistance to changes in shape, whereas when it was exposed to forces exceeding a threshold, microgels underwent permanent deformation (Figure 8E). An average of 878 Pa at 0.63 rad s⁻¹ also showed similar elasticity to the human skin (Figure S19, Supporting Information).^[18]

After creating 6 mm diameter wounds with ≈1 mm thickness on the dorsal skin, precise scanning allowed automatic generation of printing paths using pointclouds specific to each wound. Real-time scanning ensured accurate adaptation of the printing path to the slight variations in the soft skin layers caused by movements of the subject, enabling precise targeting during the procedure. On day six, both groups displayed over 50% wound closure, while a slight difference in wound healing speed between the two groups was observed on day 10 (Figure 8F,G). Masson’s Trichrome (MT) staining (Figure 8H,I) revealed well-developed granulation tissue and higher cell density within the wound area, indicating advanced healing in the microgel-treated group. It can be stated that gelatin, by inducing angiogenesis and providing a structural supportive environment for infiltrating cells, enhances the process of wound healing.^[24] Moreover, compared to the Tegaderm only treated (TD only) group, the microgel-treated wounds exhibited a significantly increased area of migrating hyperproliferative neo-epidermis, as depicted in Figure 8J which is indicative of the beneficial effect of epidermal keratinocytes.

3. Discussion

This research on robotic in situ printing stands as the first of its kind to combine path-to-motion robot control, followed by rapid and precise printing. INSIGHT is a notable advancement in the bioprinting field, going beyond the restrictions of conventional 3-axis systems by providing greater freedom in scanning and printing spatial and temporal areas. The considerable time investment in diagnosing patient-specific wounds, pre-designing printing codes for creating 3D constructs poses a hindrance to promptly printing on wounds encountered in a hospital setting. Additionally, there is a limitation in incorporating synergistic therapeutic effects into bioinks to enhance wound healing while maintaining printing capabilities.

The materials employed in this study are commonly used for 3D printing biomaterials. However, to broaden the technology’s applicability in providing initial treatment for various wounds, INSIGHT demonstrates that synchronized pneumatic valve control enables the use of different inks tailored for specific purposes. This showcase underscores the versatility of our technological capabilities across diverse application areas, including the 3D printing of natural polymer blend ink, micro-granular gel-based ink, and biocompatible elastomeric ink.

INSIGHT could pave the way for the growth of the medical imaging technology market and could lead to efficient diagnosis and treatment in clinical settings. This cutting-edge technology enables rapid emergency treatment for trauma patients and facilitates the selection and application of customized therapies. Consequently, it is crucial to prioritize the development of fundamental medical materials applicable to in situ bioprinting. These emerging technologies can be engineered with the ability to print various types of wound defects with varying depths, shapes, and sizes, thus allowing for the management of a diverse range of clinical scenarios. By establishing a remote operation-based platform for clinical experts in hospitals, this intelligent technology can benefit trauma patients while achieving significant cost savings. In addition, by controlling fine-tuning pneumatic valve gating to account for the distinct rheological properties of different inks as well as by developing various designs of the printing nozzle with the assembled bioink reservoirs, INSIGHT presents an enormous opportunity to offer a diverse range of new treatment options, particularly in the absence of effective treatment methods for traumatic and chronic wounds. Furthermore, for future work, we are developing a bioink based on microgels, incorporating human mesenchymal stem cells (hMSCs) and oxygen-generating microparticles (OMPs) to investigate the synergistic effect of combining oxygen and stem cells for diabetic wound healing. In transitioning to the clinical application phase, there are still limitations. Limitations in the acquired data may introduce uncertainties into computer vision methods, compromising the accuracy of scene reconstruction, especially concerning the limitations of the current color detection method in capturing all wound edges. This, in turn, may impact the precision of plans generated and executed by the robot. To address these issues, proposed risk mitigation strategies include the adoption of advanced convolutional neural techniques to enhance image fidelity for planning. Constraints in the data obtained can introduce inaccuracies into computer vision methods, leading to a reduction in the quality of scene reconstruction. For instance, the current color detection method may struggle to distinguish between wounds and potential blood flow, which in turn can impact the accuracy of plans generated and subsequently executed by the robot. To implement this technology effectively in real clinical scenarios, it appears that primary treatment such as blood suction is necessary.

Furthermore, in anticipation of broader clinical use, fully automated AI systems are poised to be highly valued, offering considerable cost savings for professionals such as clinicians and data scientists by integrating vision-guided based software and hardware including robotic arm-based printing. This exceptional and high-potential technology creates the potential for expedited and customizable first-aid wound therapy in the field of regenerative therapeutics.

4. Experimental Section

Materials and Reagents:

Gelatin from porcine skin (Gel strength 300, type A), alginic acid sodium salt from brown algae (Medium viscosity), poly(ethylene oxide) (PEO, molecular weight of 300 000 g mol⁻¹), sodium chloride (NaCl), calcium chloride (CaCl₂), methacrylic anhydride (MA), glycerol, sebacic acid, 4-(dimethylamino)pyridine, anhydrous dichloromethane (DCM), acryloyl chloride, triethylamine, ethyl acetate, 2-Hydroxy-4-(2-hydroxyethoxy)–2-methylpropiophenone (Irgacure 2959) were purchased from Sigma-Aldrich (St. Louis, MO) and used as received unless otherwise reported. Dulbecco’s Phosphate Buffered Saline (1x) (DPBS) was purchased from Gibco. Cellulose Nanofibrils (CNFs), prepared by supermass colloider (Cellulose Nanofibril 3 wt% in water) was purchased from Cellulose Lab. Clear resin V4 was purchased from Formlabs.

System Setup:

An EPSON VT6L All-in-One 6-Axis robot was utilized and integrated with a depth camera (D405, Intel RealSense) and a pneumatic printing system. The depth camera was securely mounted using an M6 screw on a custom designed holder. The pneumatic printing system comprised components like an ink reservoir, nozzle, piston, and adaptor, all obtained from Integrated Dispensing Solutions, Inc. These components were assembled and loaded onto another custom-made holder. Camera and syringe holders were designed using Fusion 360 software and then fabricated using a 3D printer (Form 3, Formlabs, USA) with the clear resign. The optimized dual-mode printhead is installed in the digital multi-channel on/off switching system and is linked to pressurized tanks through tubing, which is managed by an Ethernet-enabled valve control system (WAGO Programmable Fieldbus Controller, USA).

Ink Preparation:

To demonstrate all functions of printing with the pneumatic printing system, a gelatin-based ink was used. To prepare 20 mL volume of gelatin-based ink, 0.4 g of gelatin was dissolved in 6.3 mL DPBS (1x) and stirred at 50 °C for 10 min. The 0.4 g of alginic acid sodium salt was added and stirred at 50 °C for 10 min. After then, 13.7 g of CNFs was added and mixed with a spatula to obtain a homogeneous ink. The prepared ink was aliquoted in a syringe and kept in the 4 °C fridge.

For bandaid printing, photocurable tissue adhesive, poly(glycerol-cosebacate) acrylate (PGSA), was synthesized as previously reported.^[21b] PGS prepolymer was prepared by stirring glycerol (3.7 ml, 50 mmol) with sebacic acid (10.1 g, 50 mmol) at 120 °C under nitrogen for 8 h. The pressure was subsequently reduced to 50 mmHg and the polycondensation reaction was continued for an additional 16 h at 120 °C. PGS prepolymer (14 g) and DMAP (14 mg, 0.11 mmol) were dissolved in DCM (200 mL) under nitrogen and cooled to 0 °C. Acryloyl chloride (2 mL, 25 mmol) and triethylamine (3.5 mL, 25 mmol) were added dropwise while stirring. The reaction was covered in aluminum foil, warmed to room temperature, and stirred for 24 h under nitrogen. DCM was evaporated under reduced pressure. The resulting residue was dissolved in ethyl acetate (200 mL), stirred for 15 min, and filtered to remove the precipitates. Ethyl acetate was removed under reduced pressure. To PGSA solution in DCM (50 w/v%), 0.25 w/v% photoinitiator Irgacure 2959 was added, and DCM was evaporated before use as the ink for bandaid printing.

For in vivo experiments, the microgels were prepared using the combination of gelatin and gelatin methacryloyl (GelMA) through water in water emulsion method according to the methodology proposed by Ying et al. with slight modifications.^[22a,25] GelMA synthesis was performed as reported previously.^[26] Briefly, 10% (w/v) gelatin was dissolved in a sterile PBS while stirring continuously at 50 °C with a rotation of 240 rpm for 30 min until a homogenous consistency was seen. Conditions of agitation and temperature were maintained while adding 0.5 mL of MA in a continuous fashion. Continuous stirring was applied at a temperature of 50 °C for two hours, after which the reaction was stopped by addition of 200 mL prewarmed PBS. This solution was dialyzed in deionized water for 5 days using 12–14 kDa dialysis membranes to remove unreacted MA. Finally, the solution was filtered using a 0.22 μm pore size of a filtration cup and lyophilized for 5 days. Degree of methacrylation (DoM) was determined using OPA. Briefly, GelMA was dissolved in PBS at 0.5 mg mL⁻¹ and mixed in a 1:1 volume ratio with OPA. Gelatin at 0.5 mg mL⁻¹ functions as a standard and PBS as a blank. 200 μL of these mixtures was added to a 96 well plate in triplicate and fluorescence intensity was read using the microplate reader at 340 nm/455 nm. After this, the DoM was calculated using the following formula:

$$DoM(\%)=\frac{I_{sample}-I_{sample}}{I_{s\tan dard}-I_{blank}}$$ (1)

For the microgel preparation, a low degree of methacrylation of GelMA, around 30% was used. Briefly, 125 mg of gelatin and 125 mg of GelMA with around 30% of DoM were dissolved in 2.5 mL of distilled water (DI). Additionally, 90 mg of NaCl was dissolved in 2 mL of DI. All three mixtures were set into a hot water bath at 50–60 °C for 25 min and mixed together by vortexing for 30 s and set in a water bath at 50 °C for 5 min. Once dissolved, the mixture was added to the PEO solution and vortexed until an opaque white thick consistency was seen. The mixture was left on ice for 20 min. Subsequently, the mixture was centrifuged for 15 min at 4 °C under 4000 rpm for 20 min and the supernatant was removed. tubes. 10% (v/v) CaCl2 was added and subsequently DI was added in a 1:1 volume ratio. The solutions were vortexed for 30 s and centrifuged at 4 °C under 4000 rpm for 20 min. Finally, the supernatant was removed from the solutions and stored at −4 °C until their use.

Viscosity Measurement:

Viscosity of all inks was performed using a rheometer (HR-3, TA Instruments, USA). 200 μl of inks were loaded for analysis on the 20 mm diameter flat plate of the rheometer and their viscosity was measured in the range of shear rate 0.01 to 100 s⁻¹. and as a function of frequency in the range of 0.1 to 100 Hz at 0.1% strain, and as a function of strain in the range of 0.1 to 100% at 1 Hz. at room temperature. The recovery behavior of the ink was carried out by a step-strain sweep where low strain (1%) and high strain (90%) were cycled every 100 s with a frequency of 1 Hz according to previous method.^[27] The viscosity of polymer blend ink was also measured in a temperature sweep from 0 to 45 °C. The viscosity of the bandaid ink was also obtained in strain ratecontrolled measurements with 0.1% strain and shear rates from 0.1 to 100 s⁻¹, and its strain sweep tests were performed with a frequency of 1 Hz in the range of 0.1 to 100%, and frequency sweep tests were conducted with a strain of 0.1% in the range of 0.1 to 100 Hz. For the microgel inks, viscosity test was conducted at 4 °C.

Rheological Behavior Measurement:

Storage and loss modulus of all crosslinked inks were measured using a rheometer (HR-3, TA Instruments, USA). Samples were loaded for analysis on the 20 mm diameter flat plate of the rheometer and their storage and loss modulus were measured as a function of frequency in the range of 0.1 to 100 Hz at 0.1% strain.

Ex Vivo Experiments:

ex vivo experiments were performed using rodent and porcine models. For ex vivo rodent models, 300 day-old- Sprague-Dawley (SD) rats were purchased with the approval of Brigham Women’s Hospital (BWH) Institutional Animal Care and Use Committee (IACUC) (IACUC #2017N000114). After sacrifice, various shapes of wounds were prepared using biopsy punch and surgical scissors. For ex vivo porcine models, yorkshire pig round 40 kg was sacrificed and various shapes of wounds were prepared using surgical scissors (Tissue only protocol IACUC #2022N000194).

Spray Simulations:

The numerical simulation of spraying inks was conducted using air-blast atomizer model and detailed simulation process is depicted in Supporting Information.

In Vivo Diabetic Models:

All procedures were conducted in accordance IACUC, under protocol 062–2021. 12-week-old male db/db mice (strain 000642) were obtained from The Jackson Laboratory. Their hyperglycemia was confirmed by blood glucose levels exceeding 250 mg dL⁻¹. After anesthetized using isoflurane, and) full-thickness wounds were created on their depilated and disinfected dorsum using two circular 6 mm biopsy punches (Integra Miltex). After wound recognition using INSIGHT, microgels were directly deposited on the wound and photo-crosslinked using UV light for 30 s. The wounds were then covered with an occlusive dressing (Tegaderm, 3 M) for protection. The mice were accommodated in individual housing and subjected to daily observation until they were euthanized using an excess of CO₂. The wounds were monitored on days 0, 3, 6, and 10, photographed and measured with digital calipers and wound closure was presented as percentage healed compared to day 0. At Day 10, mice were sacrificed and all wound tissues were bisected at the wound center, and fixed in 10% formalin. The tissue samples were prepared for paraffin embedding. Sections of 5 μm thickness were used. Hematoxylin and eosin (H&E) and Masson’s trichrome (MT) staining was performed accordingly. Stained images were acquired via a slide scanner (Zeiss AxioScan7).

Statistical Analysis:

Data analysis was performed by using GraphPad Prism 9 (GraphPad Software Inc., USA). The values represent the mean ± SD from three or more independent experiments. The statistical significance of the differences was determined by one-way analysis of variance (ANOVA). * p < 0.05, and *** p < 0.001 were considered statistically significant.

Supplementary Material

Supporting Information

Supporting Information is available from the Wiley Online Library or from the author.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.