This study used an experimental, mixed methods design to develop, characterize, and evaluate Flexoknit, a transtibial prosthetic interface created through CNC multimaterial knitting and informed by residual limb skin strain patterns. The study comprised three major phases: (1) digital design and fabrication of multimaterial knitted structures, including yarn selection, stitch programming, and interface construction; (2) material and fabric characterization involving thermal analysis, mechanical testing, and moisture management assessment of the knitted samples; and (3) user testing to evaluate mobility, suspension, thermal and moisture comfort, and overall user satisfaction compared with a conventional silicone liner.

This design enabled a comprehensive evaluation of both the engineered textile properties and the user-level functional response to a digitally informed prosthetic interface.

Participants were recruited from a prosthetics clinic in Singapore using convenience sampling. Inclusion criteria included participants aged ≥21 years, unilateral transtibial amputation, regular prosthesis users, and the ability to ambulate independently with or without assistive devices. Exclusion criteria included active dermatological issues on the residual limb, acute musculoskeletal conditions, or medical conditions affecting ambulation.

One long-term transtibial prosthesis user (76-y-old man) completed the full user testing protocol, including mobility, suspension, thermal, moisture, and satisfaction assessments. An additional healthy participant (35-y-old man) performed the outdoor thermal test in hot and humid conditions to simulate typical Singapore climate exposure. The difference in participants between the indoor and outdoor tests was due to the age of the amputee participant. As the intention of the outdoor sweat and thermal test was to generate adequate sweat within a relatively short duration (15 min) by having a user who could exert themselves under the hot sun, the team opted for a healthy younger participant to perform this outdoor test instead.

For the main body of the Flexoknit interface, 2 types of thermal-reactive yarns were used. TH1 consists of a 10/1 bicomponent polyester binder spun yarn supplied by Unitika, branded under the model ESPORAN. This yarn is constructed with a polyester core that has a melting point of 256 °C and a co-polyester sheath with a melting point of 110 °C. TH2 is made from a blend of polyester, polyurethane, and nylon, featuring a fusing point of 100 °C, supplied by Asahi Kasei. Three types of spandex yarns were used: ACY (300 D polyester DTY (96 F) and 100 D spandex); ZJS (210 D spandex and 2 ends of 75 D polyester, under the model number 2107575) from Zhejiang Kangjiesi New Material Technology Co., Ltd; and SPY, a low-stiffness elastic yarn from Shima Seiki called Shima Set Up Yarn.

The yarns were knitted into fabric on a Shima Seiki MACH2XS153-15L WHOLEGARMENT 15-gauge flat-bed weft knitting machine programmed using the SDS-ONE KnitPaint software. Three stitch patterns were used in the knitted prosthetic interface (see Figure 1 for diagrams of the stitch patterns). Single jersey (SJ), also known as plain jersey or stockinette stitch, is formed with only front knit stitches. SJ with inlay (SJ-IL) consists of 1 yarn to knit a SJ fabric, while a second yarn is woven in and out of the knitted loops of the first yarn. The second yarn forms approximately straight floats of yarn along the course of the fabric. Knit-miss (cross-miss) 1×1 with tucks (KM-TK) uses 1 yarn to knit the knit-miss 1×1 stitch pattern, where alternate needles knit and miss stitches, with the order of the knit and miss stitches reversing in every course. A second yarn tucks on the knit-miss 1×1 fabric at regular intervals.

More information on the fabrication of the prosthesis can be found in Multimedia Appendix 1.

LoNEs are regions of the body that undergo negligible skin strain during movement [27]. Prior studies have suggested using LoNEs to guide material placement in wearable systems—including prosthetic sockets [33] and mechanical counterpressure suits [34]—by positioning stiff materials along low-strain areas and reserving compliant materials for regions of higher deformation. This strategy allows structural reinforcement without restricting natural skin mobility. Similar approaches have been applied in orthotic design using selectively reinforced 3D-printed elements [35].

In this study, LoNEs informed the spatial layout of stiffness zones within the knitted prosthetic interface. The user’s residual limb was photographed in multiple poses, reconstructed as a 3D model, and analyzed to generate directional strain fields (Figure 2A; see [28] for full methodology). High-, medium-, and low-stiffness zones were then mapped based on these strain patterns (Figure 2B), forming the structural template for the knitted design. In this study, mapping was conducted under controlled, non–weight-bearing conditions, with the participant supported and not loading the residual limb. Skin deformation was captured between 2 standardized knee positions: full extension (0°) and flexion at approximately 60°, measured using a goniometer to ensure consistency.

For each posture, a circumferential video of the residual limb was recorded over approximately 90 seconds, from which a sequence of images (~300 frames) was extracted for 3D reconstruction. The strain field was computed by comparing the reconstructed limb geometry between the extended and flexed states, and the resulting directional strain data were used to identify LoNEs for interface design. A single mapping sequence was performed for the participant, and the derived strain field was used directly for design implementation.

This mapping approach captures kinematic skin strain behavior under controlled conditions and does not account for additional deformation under load-bearing scenarios. Accordingly, LoNEs are interpreted as functionally low-strain design pathways rather than fixed anatomical lines under dynamic loading.

High-stiffness zones were positioned directly over LoNE pathways, providing mechanical reinforcement without restricting mobility. These areas resist elongation under load, supporting suspension, minimizing pistoning, and redistributing pressure toward more tolerant regions of the residuum, such as the medial and lateral tibial surfaces and calf musculature [36]. A stiffened distal zone was added to improve load transfer at the socket-interface junction while reducing overall weight by confining rigid materials only to targeted regions.

Medium-stiffness zones were placed adjacent to the patella, where controlled flexibility is required to accommodate knee movement. This configuration provided directional stiffness in the vertical axis to prevent interface deformation and limit migration while preserving transverse flexibility for knee flexion.

Low-stiffness zones corresponded to regions experiencing high skin strain and were designed to stretch freely, supporting soft tissue deformation during movement. As these zones form most of the interface surface, using breathable, compliant yarns in these areas substantially improved comfort and thermal and moisture management. Integrated compression structures in both layers further improved mechanical coupling between limb and interface and helped manage shear.

The final Flexoknit interface was constructed as a double-layered knitted tube (Figure 3). The outer layer incorporated spatially graded stiffness elements through specific yarn and stitch combinations. High-stiffness zones were primarily knitted using the KM-TK_TH1-ACY configuration, while SJ_TH1 was used at the distal end to allow 3D shaping without compromising rigidity. Medium-stiffness zones used KM-TK_TH2-ACY, and lateral and medial regions included integrally knitted channels containing spiral steel boning. The boning provided vertical stiffness to reduce pistoning while preserving transverse flexibility for knee movement.

Low-stiffness regions were knitted with compliant spandex yarns (eg, SJ_ACY) to enable close limb conformity and accommodate local strain. Compression was strategically integrated: SJ-IL_ACY-SPY provided CW compression in the outer layer, while SJ-IL_ACY-ZJS applied surface-level compression in the inner layer, enhancing mechanical coupling and reducing shear.

Additional design features addressed impact protection, suspension, and adjustability. Cushioning pads were placed over vulnerable anatomical areas, including the fibular head, tibial crest, and distal stump. Suspension was facilitated by the thermoformed fit of the knitted tube, augmented by hook-and-loop straps, a patellar silicone ring, and antislip regions aligned with LoNEs. These elements improved security during ambulation without compromising donning ease or adaptability to limb volume fluctuations.

The knitted prosthetic interface was fabricated on a Shima Seiki MACH2XS153-15L WHOLEGARMENT 15-gauge flat-bed weft knitting machine, with all stitch programming executed in SDS-ONE APEX/KnitPaint. Thermal-reactive yarns (TH1 and TH2) were used to enable structural fixation after heat setting. During the user tests, the knitted prosthetic interface was inserted into a rigid laminated transtibial socket and aligned to ensure correct placement of stiffness zones.

Following fabrication of the Flexoknit prosthetic interface, a series of controlled tests was conducted to compare performance against the participant’s existing prosthesis with a conventional silicone liner. User testing was organized into 3 domains (ie, mobility, suspension, and comfort) using standardized quantitative assessments and structured qualitative data collection. A summary of the test protocol is provided in Table S2 in Multimedia Appendix 1.

A static elongation test was carried out. Two reference marks were applied along the proximal brim of the rigid socket—one at the highest lateral point and the other aligned with the patella. With the participant standing, the prosthetic limb was lifted while maintaining full knee extension. Vertical displacement between the original and displaced marks was measured to determine socket migration relative to the knitted interface. Displacement was measured manually using a ruler aligned to the reference markings, with a measurement resolution of 1 mm, and measurements were repeated across 3 trials for each prosthetic condition, with the mean value reported. The measurement error for vertical displacement using manual reference markings was estimated to be +1 or −1 mm based on repeated pilot measurements. Any measured variability across the 3 trials would also be reported as well.

Testing was performed with the knee maintained in full extension throughout to minimize variability due to joint position. Limb loading was standardized using the participant’s body weight during standing prior to limb lifting, followed by manual unloading of the prosthesis; however, no instrumented load measurement was used, and the applied force during limb lifting was not quantified.

This method represents a static assessment of suspension and does not account for dynamic loading conditions encountered during gait, including cyclic loading, shear forces, and stance-phase transitions. Given the manual measurement resolution and estimated error margin, small between-condition differences should be interpreted with caution. Supplementary feedback on suspension was collected through brief unstructured interviews.

Material properties (DSC peaks, breaking force, strain at break, modulus, thermal resistance, vapor permeability, and wicking rate) were analyzed. Tensile results were plotted as force-strain curves. Mobility and suspension outcomes were compared between the Flexoknit interface and the participant’s silicone liner. Temperature data were analyzed as continuous time series profiles. Moisture comfort and qualitative feedback were thematically summarized. Questionnaire scores (SCS and QUEST 2.0) were tabulated and compared across configurations.

This study comprised a material characterization phase and a single-user proof-of-concept evaluation. Owing to the single participant design of the user testing component (n=1), no inferential statistical analysis was performed for the clinical, functional, thermal, or satisfaction outcomes. Normality testing and hypothesis testing were not appropriate for single participant data. User testing results are therefore presented descriptively as raw values and percentage differences between the experimental and conventional prosthetic conditions.

For specific outcome measures, ROM, gait, mass, thermal comfort, and SCS data are reported as observed values, while QUEST 2.0 results are summarized using median scores. Suspension testing using the static elongation method is reported descriptively, with mean displacement values presented for each condition.

Material and fabric characterization data were also analyzed descriptively. Thermal and moisture management results are presented as mean values, as presented in Table 1, and mechanical behavior is presented using observed ranges and representative force-strain responses across yarn-stitch combinations. As this study was designed as a proof-of-concept investigation, no inferential comparisons were performed.

This study was approved by the Institutional Review Board of the Singapore University of Technology and Design (HBR-22‐00507). All participants provided written informed consent prior to participation. Participants were informed of the study objectives, procedures, potential risks, and benefits, and were free to withdraw from the study at any time without penalty. All data were deidentified prior to analysis and reporting to protect participant privacy and confidentiality. Only the study investigators had access to identifiable participant information, which was stored securely in accordance with institutional data protection requirements. Participants did not receive financial compensation for participation in this study.

One transtibial prosthesis user (76-y-old man, height 173 cm, weight 78 kg, and 15 y post amputation due to diabetes) completed the full mobility, suspension, comfort, and user experience testing. The participant’s conventional prosthesis consisted of a 6 mm silicone locking liner, laminated composite socket, pylon components with a pin lock suspension system, and a 1C30 Trias foot, worn for an average of 9 hours daily.

Socket migration measured using the static elongation test showed a mean displacement of 4 mm (+1 or −1 mm) with the experimental system and 3 mm (+1 or −1 mm) with the conventional system. Given that the observed between-condition difference (1 mm) lies within the estimated measurement error of the manual marking method (+1 or −1 mm), this finding should be interpreted with caution. During indoor walking, the participant reported a gradual loosening sensation and sinking at the distal tibia when using the experimental prosthesis, despite the small measured difference in static displacement. These observations were recorded during the structured suspension assessments.

This proof-of-concept study demonstrates the feasibility of a digitally informed, multimaterial knitted prosthetic interface for transtibial users while also highlighting the biomechanical trade-offs introduced by increased interface compliance. By combining residual limb skin strain analysis with CNC multimaterial knitting, we created Flexoknit—a double-layered, graded stiffness interface integrated with a LoNE-informed rigid socket. Material characterization showed that targeted yarn-stitch combinations can generate 3 distinct stiffness tiers spanning nearly an order of magnitude in elastic modulus. When implemented in a transtibial system, this graded design reduced interface mass relative to a conventional silicone liner, preserved or improved knee ROM, and showed context-dependent thermal and moisture advantages, particularly in hot outdoor conditions. However, suspension-related challenges, including the participant’s reported distal “sinking” sensation, and the reduced distance walked during the 10MinWT indicate that improvements in ROM and comfort-related properties do not necessarily translate into better gait performance or interface stability. Despite the suspension-related challenges and donning complexity, user-reported satisfaction scores and qualitative feedback indicate that the concept is acceptable and potentially attractive for lower-activity community amputees. Taken together, these findings suggest that increased interface flexibility may have facilitated greater knee excursion during isolated ROM testing, while simultaneously reducing stance-phase stability during walking through less consistent limb-socket coupling. These findings therefore support feasibility but also highlight the need for further refinement of load transfer and suspension security before broader clinical application.

The material characterization phase confirms that both yarn composition and stitch architecture can be systematically tuned to control mechanical behavior. Thermal-reactive yarns (TH1 and TH2) exhibited clear melting transitions and, once heat-set, produced knitted structures with substantially higher stiffness and strength than fabrics knitted solely from spandex yarns, consistent with prior work on thermal-reactive knits [20]. Uniaxial tensile testing showed that KM-TK_TH1-ACY and SJ_TH1 formed the high-stiffness group, KM-TK_TH2-ACY formed the medium-stiffness group, and SJ spandex configurations formed the low-stiffness group. The separation between these categories—38‐288 N/m; 1224‐2978 N/m and 12,886‐110,337 N/m, respectively—provided a clear design palette for allocating support and compliance within the interface.

Prestretching the knitted tubes prior to heat setting, a process that mimics forming the fabric over a limb mold, further modified the mechanical response. Narrow (stretched) samples typically showed increased stiffness and breaking force in the CW direction and more homogeneous behavior between WW and CW axes, suggesting reduced anisotropy once the fabric is tensioned to its in-use configuration. This behavior is advantageous for prosthetic applications, where a predictable, direction-independent response can improve suspension and load transfer.

Thermal and moisture management testing indicated that stitch patterns can be used to optimize comfort-related properties independently of stiffness. Among the patterns assessed, SJ achieved the best overall balance of low thermal resistance, high water vapor permeability, and adequate vertical wicking, supporting its use as the dominant structure in low-stiffness, high-contact zones. Collectively, these findings show that multimaterial CNC knitting can deliver a controlled spectrum of mechanical, thermal, and moisture management properties within a single seamless interface.

The LoNE-guided, digitally manufactured construct illustrates a shift from uniform material liners toward spatially tailored interfaces in which stiffness, compression, and breathability are all driven by patient-specific biomechanical data. LoNE-guided reinforcement may improve suspension by placing stiffer load-bearing elements along relatively low-strain pathways, where they can resist migration with less disruption to natural tissue motion. However, LoNEs are not assumed to remain perfectly fixed under load-bearing conditions and should instead be interpreted as functionally stable design zones. Nevertheless, the present findings suggest that anatomical alignment of stiffness zones alone is insufficient; effective suspension depends not only on alignment with LoNE pathways but also on the absolute magnitude and spatial distribution of interface stiffness required to maintain stable load transfer during dynamic gait.

In functional testing with the participant, the Flexoknit system preserved maximum knee extension and increased overall knee flexion when compared with the conventional prosthesis, yielding a 22.5% improvement in total ROM. This suggests that aligning stiff zones with LoNEs and reserving compliant structures for high-strain regions may permit greater sagittal-plane excursion at the knee without immediately compromising basic structural support. However, the increase in ROM should not be interpreted as a direct proxy for improved functional mobility, as gait performance during the 10MinWT, 10mWT, and TUG was modestly worse with the experimental system. Despite the 26.1% reduction in prosthesis mass, the participant walked more slowly with the experimental system. This suggests that prosthetic weight alone was not the dominant determinant of performance in this case. From a biomechanical perspective, walking efficiency depends not only on limb mass but also on the quality of load transfer, proprioceptive consistency, and the user’s confidence during the stance phase. It is therefore plausible that the benefits of reduced mass were offset by less stable mechanical coupling at the limb-socket interface. Importantly, the participant was able to ambulate independently and perform functional tasks such as sit-to-stand, indicating that the experimental system met minimum requirements for community mobility.

The participant’s report of a gradual distal tibial “sinking” sensation is biomechanically significant, suggesting incomplete load transfer coupling between the residual limb, knitted interface, and rigid socket. In an optimally coupled system, axial and shear forces are distributed proximally and circumferentially to maintain stable limb positioning. In this case, the increased compliance of the knitted interface likely permitted localized deformation—particularly distally—reducing the efficiency of axial load transfer. This may explain the perceived distal settling despite only a 1 mm difference in static elongation, highlighting that small displacements can be clinically meaningful when repeatedly experienced at sensitive anatomical sites.

Although the study was not designed to isolate the precise structural cause, this behavior likely reflects a combination of compliant material response and insufficient longitudinal reinforcement in low-stiffness regions. While these compliant zones were intended to accommodate soft tissue and enhance comfort, they may have allowed excessive vertical deformation under load. Future iterations should therefore refine not only overall stiffness but also the spatial distribution of reinforcement to improve suspension without compromising comfort.

This phenomenon also has implications for gait stability. While compliance can enhance pressure distribution, excessive deformation may introduce micromovements that reduce proprioceptive consistency and perceived security during stance. This may contribute to a more cautious gait pattern, reduced efficiency, and slower functional performance, even in the absence of pain. The participant’s slower walking speed and longer TUG time are therefore likely influenced not only by device unfamiliarity but also by altered mechanical coupling at the interface, potentially reducing confidence during weight acceptance and forward progression.

The graded stiffness design also achieved a substantial reduction in mass, with the experimental prosthesis approximately one-quarter lighter overall and the knitted interface over one-third lighter than the silicone liner. The participant’s perception of the prosthesis as “lighter” and “springy” aligns with these findings and may explain the increased knee flexion observed during ROM testing. However, this reduction in mass did not translate to faster gait. The perceived “springiness” likely reflects increased compliance, whereby some mechanical energy during stance was absorbed within the interface rather than efficiently transferred to forward progression. While this may enhance comfort, it may also reduce the stability required for confident movement. These findings suggest that reductions in prosthetic weight should be considered alongside interface stability, as improvements in one may be offset by compromises in the other.

Thermal and moisture outcomes were environment dependent. In the air-conditioned clinic, the knitted interface produced slightly greater temperature rises than the silicone liner, suggesting increased insulation when ambient temperatures are low. In contrast, in a hot outdoor setting representative of tropical climates, the knitted interface limited skin temperature rise and left the skin dry after walking, while the silicone liner trapped sweat and was perceived as hot and sticky. These findings imply that an open, wicking knitted structure may offer particular advantages in warm, humid environments, where dermatological complications and heat intolerance are common.

Despite generally adequate suspension, the static elongation test and user reports indicated minor distal “sinking” with the experimental system. This finding likely reflects the combined mechanical behavior of the compliant knitted interface, the stiffness zoning strategy, and the structural response of the LoNE-informed rigid socket under dynamic load. Rather than representing gross pistoning alone, the sensation may indicate localized compression and delayed load sharing across the limb-liner-socket complex, particularly during repeated stance loading. Clinically, this is relevant because even subtle distal displacement can alter pressure concentration, disturb the user’s perception of stability, and reduce confidence during walking. This observation may also help explain the slower walking speed and longer TUG duration despite the reduced prosthetic mass. Even subtle distal settling or micromovement at the interface can alter how ground reaction forces are transmitted through the prosthesis and may reduce the user’s confidence during loading response and mid-stance. In response, the user may adopt a more cautious gait strategy, with slower progression and more conservative transfer during transitional movements such as turning and sit-to-stand. This likely reflects the combined behavior of the compliant interface and the LoNE-informed rigid socket, including possible flexing of the latter under load.

The lower SCS for Flexoknit compared with the user’s own device appears to be driven by this sensation of gradual loosening rather than by gross discomfort or pain. This interpretation is also consistent with the higher overall QUEST score and positive qualitative feedback regarding appearance, weight, and perceived compliance, suggesting that user acceptance of the concept remained favorable despite the mechanical shortcoming. Taken together, these findings suggest that interface compliance must be carefully balanced: sufficient to accommodate tissue motion and improve comfort, but not so great that it compromises distal support or mechanical coupling.

Future iterations should therefore investigate increasing distal and longitudinal stiffness, refining the suspension mechanism, and optimizing outer socket rigidity to reduce perceptible settling during gait. Nonetheless, the higher overall QUEST score and the participant’s preference for the device’s appearance and perceived security suggest that, with refinements to the outer socket stiffness and suspension strategy, user comfort could be further improved.

This study presents a proof-of-concept transtibial prosthetic interface—Flexoknit—engineered through a multimaterial, graded stiffness knitting approach informed by LoNEs. By combining thermal-reactive yarns, spandex-based elastic yarns, and spatially targeted stitch architectures, the knitted interface achieved distinct zones of mechanical behavior that were aligned with the user’s residual limb strain patterns and functional requirements. The fabrication process demonstrated that CNC knitting can reliably integrate structural reinforcement, compliance, compression, and breathability within a single seamless textile construct tailored to an amputee’s limb geometry.

Material testing confirmed that the yarn-stitch combinations produced the intended stiffness gradients, while functional testing showed that the interface supported safe ambulation and preserved or enhanced joint mobility when used with a LoNE-aligned rigid socket. Improvements in thermal and moisture regulation, particularly in warm outdoor conditions, and a substantial reduction in overall prosthesis weight further demonstrated the potential benefits of integrating knitted structures into prosthetic design. User feedback highlighted the comfort, lightness, and aesthetic appeal of the device, reinforcing the feasibility of this approach for improving user experience.

As an early prototype, some usability challenges remain, including potentially inadequate prosthesis stability under load and the need for more intuitive suspension and donning mechanisms. Nonetheless, the findings suggest that digitally engineered knitted interfaces can provide a highly customizable, breathable, and compliant alternative to conventional silicone liners, particularly for lower-activity amputees or individuals prioritizing comfort and ease of use.

Overall, this work establishes Flexoknit as a promising direction for future prosthetic development—one that integrates principles of biomechanics, textile engineering, and digital fabrication to create user-centered interface solutions. Further research is warranted to refine structural performance, evaluate long-term durability, and validate the design across diverse user groups and real-world conditions.

This study has several limitations that should be considered when interpreting the findings. First, the prosthetic evaluation was conducted with a single transtibial amputee user, and thermal testing in a hot outdoor environment involved only one additional healthy participant. The results therefore reflect an in-depth case study rather than population-level evidence and may not generalize to users with different limb geometries, activity levels, or comorbidities. Second, the familiarization period with the Flexoknit system was short, so the gait and comfort outcomes likely include early adaptation effects. Longer-term use may alter both objective performance and subjective preferences. Third, the prototype system combined a novel knitted interface with a bespoke LoNE-informed rigid socket that has not yet been fully optimized. The biomechanical interpretation of distal “sinking” is based on user reports, static elongation data, and observational findings rather than instrumented assessment of in-socket motion or pressure distribution. In addition, while material and bench-top testing were extensive, real-world durability, laundering behavior, and the long-term skin response to continuous fabric–skin contact were not assessed. These aspects are critical for translation to everyday clinical practice.

Future work should therefore focus on multiparticipant studies that include a broader range of ages, activity classifications, and residual limb shapes, with follow-up over weeks to months. Such studies should incorporate more detailed biomechanical outcomes (eg, 3D gait analysis, interface pressure, and shear mapping) and dermatological assessments to evaluate skin integrity over time. On the design side, iterative optimization of the rigid socket stiffness, strap layout, and donning sequence is needed to reduce distal sinking and simplify daily use. Finally, further development of the digital workflow—from automated LoNE extraction to direct translation into knit patterns—along with systematic durability and wash-cycle testing, will be essential to move Flexoknit from proof-of-concept to a clinically deployable, manufacturable system.