A Closed-Loop Digital–Physical Workflow for Patient-Specific Surgical Guides in Maxillofacial Reconstruction: A Clinical Feasibility Study

Abstract

This study presents a closed-loop digital–physical workflow for manufacturing patient-specific anatomical models and surgical guides for maxillofacial reconstruction. The workflow combines medical-image segmentation, additive manufacturing of anatomical models, surgeon-led physical simulation, re-digitization of the modified models, landmark-based and surface-based alignment with the original CT-derived anatomy, with FFD used as a limited supporting approach for localized geometry updating, CAD design, and fabrication of sterilizable guides. Technical feasibility was examined in two detailed mandibular reconstruction cases involving temporomandibular joint prosthesis implantation. The workflow was subsequently used in adapted forms in eight clinically heterogeneous mandibular reconstruction cases related to oncological treatment. These cases are reported as supplementary clinical implementations and were neither treated as direct replications of the complete closed-loop methodology nor pooled for quantitative comparison. In the two detailed cases, the guides were verified on the anatomical models, used intraoperatively, and followed by postoperative imaging providing qualitative clinical confirmation of implant positioning. The study did not include a direct CT-based control workflow or standardized geometric deviation analysis. Therefore, operating-time and cost effects are reported as case-specific observations and an illustrative scenario rather than comparative evidence. The main contribution is the physical-to-digital feedback step, which allows surgeon-introduced modifications of a printed model to be transferred back to the digital geometry before final guide design. The results support the technical feasibility and adaptability of the workflow, while prospective quantitative validation remains necessary.

1. Introduction

Additive manufacturing methods have become essential tools in maxillofacial surgery, particularly for preoperative and intraoperative planning. They enable the creation of precise anatomical models and surgical guides, as well as implants tailored to the individual needs of the patient, providing numerous benefits. These include reduced operative time, increased surgical accuracy, and decreased costs associated with the use of cadaveric models for educational purposes, as additively manufactured models offer a more cost-effective alternative [1,2,3,4]. The most commonly used methods in maxillofacial surgery include virtual surgical planning, the production of anatomical models, and the creation of patient-specific surgical guides. Virtual planning allows for the simulation of osteotomy cuts and manipulation of bone segments, enhancing surgical accuracy and patient safety. Three-dimensional models enable the preoperative fitting of reconstruction plates, the execution of surgical simulations, and improved communication between the surgeon and the patient, thereby reducing the risk of complications. Printed surgical guides accurately replicate the planned cuts and ensure precise instrument positioning, supporting effective and personalized procedures [5,6,7,8,9]. The core methodology involves the analysis and segmentation of data from computed tomography or magnetic resonance imaging, enabling virtual simulations, including the planning of surgical cuts and reconstructions [8,10,11]. The prepared digital model is then manufactured using various methods, such as fused deposition modeling (FDM), selective laser melting (SLM), stereolithography (SLA), or PolyJet printing. These methods allow for the rapid, accurate, and cost-effective production of models that support maxillofacial surgical procedures [7,8,11,12,13,14,15,16,17,18,19].

The term point of care manufacturing (POCM) increasingly refers to the production of surgical models and guides directly within healthcare facilities using additive manufacturing technologies. This local production enables dynamic modifications of designs in response to the patient’s specific anatomical features and current clinical condition, without relying on external laboratories. Consequently, surgical teams can rapidly obtain instruments and implants tailored to individual cases, facilitating the optimization of procedures and faster implementation of innovative methods. The main advantages of POCM include reduced production time, lower costs, and improved clinical outcomes, as reported in multiple studies in the field of maxillofacial surgery. Implementation of these processes in hospitals also allows for quick adaptation to urgent cases, such as trauma or complex bony reconstructions [6,8,20,21,22,23,24].

The design of models and surgical guides starts with a digital representation of the patient’s anatomical structures, which, after segmentation of medical images, takes the form of a three-dimensional reconstruction. This model serves as the foundation for developing guides that accurately translate virtual surgical plans into the operating room. The use of this approach facilitates precise replication of planned bone cuts and positions, reducing the risk of intraoperative errors and supporting predictable and reproducible treatment outcomes [6,25,26].

The prepared models are imported in one second step into specialized software for precise surgical planning, including the definition of osteotomy lines, resection areas, and the positioning of bone segments. They also serve as a basis for designing surgical guides in CAD software. This process involves specifying cutting planes, drilling points, and the arrangement of bone segments. Models are saved in STL format, allowing them to be 3D printed. The printed models are evaluated by clinicians for accuracy, after which various plates are adjusted and fitted, and can subsequently be digitized (e.g., 3D scanning or computed tomography) to design patient-specific surgical guides. Such methodologies enable precise guide design, contributing to improved surgical outcomes and enhanced patient safety [25,27,28,29,30,31,32,33].

Virtual surgical planning, in-house additive manufacturing of anatomical models, and patient-specific shell-type surgical guides are established elements of contemporary maxillofacial reconstruction. In this study, these components are integrated into a closed-loop digital–physical workflow. Its distinctive element is the physical-to-digital feedback step following surgeon-led simulation. In the proposed workflow, the surgeon-adjusted physical model is re-digitized, and the original CT-derived digital anatomy is updated before the final guide is designed. This allows physical modifications introduced during planning to be incorporated into the digital model without repeating the complete segmentation procedure.

The aim of this study was to describe the closed-loop workflow and assess its technical feasibility and clinical integration. Two complex mandibular reconstruction cases involving temporomandibular joint prosthesis implantation are presented in detail. In addition, the workflow was used in adapted forms in eight clinically heterogeneous mandibular reconstruction cases related to oncological treatment. These cases are summarized as supplementary clinical implementations and should not be interpreted as direct replications of the complete closed-loop methodology. Because the study did not include a matched control workflow or standardized geometric deviation analysis, it was not designed to establish quantitative superiority, general clinical effectiveness, or formal cost-effectiveness.

2. Materials and Methods

2.1. Concept and Plan of the Work

The process of developing anatomical models begins with the analysis of a patient’s medical imaging obtained through computed tomography (CT). At this stage, it is important not only to accurately interpret the imaging data but also to conduct consultations with the medical team, aimed at defining the key requirements for the models being prepared. Consideration is given to aspects such as the intended clinical application, expected appearance, durability, and dimensions. Based on the collected information, the engineer proceeds with the image segmentation process, employing both manual and automated techniques, which enables the creation of a three-dimensional anatomical model. This model, after appropriate digital processing, is converted into STL format and prepared for 3D printing.

The printed model serves as a valuable tool to support surgical planning, enabling physicians to perform operative simulations, including the determination of bone resection lines and the selection and preliminary fitting of a titanium reconstruction plate. During this stage, surgeons may introduce geometry-altering modifications, such as mandibular repositioning, wax stabilization, and manual adaptation of reconstruction plates, resulting in a physical model that reflects the final intra-operative plan.

In subsequent steps, the physically adjusted model is re-digitized and aligned with the original CT-derived digital anatomy in GOM Inspect 2017. Initial manual alignment is performed using three corresponding anatomical landmarks, the mandibular condyle, the mandibular foramen at the entrance to the mandibular canal, and the mandibular angle, followed by surface-based best-fit registration using stable anatomical regions. FFD is considered as a limited supporting approach for localized geometry updating rather than as a fully parameterized deformation procedure. The aligned CT-derived and re-digitized geometries are subsequently used as complementary references during CAD design of the patient-specific surgical guides. The completed guides are verified on the corresponding anatomical models before sterilization and intraoperative use.

The workflow was implemented in collaboration with the Clinical Department of Maxillofacial Surgery at the University Clinical Hospital in Poznań. Engineering and clinical decisions were made iteratively: the clinical team specified the operative requirements and approved the physical simulation, whereas the engineering team performed image processing, model preparation, re-digitization, geometry updating, CAD design, and additive manufacturing. To facilitate understanding of the individual steps, a schematic of the process was developed, shown in Figure 1.

2.2. Overview Concerning the Cases of Patients Considered for the Research

The study was designed as an exploratory assessment of clinical feasibility. Two detailed cases involving complex mandibular reconstruction with temporomandibular joint prosthesis implantation were documented in detail because they demonstrated the complete closed-loop workflow. In addition, the workflow was applied in adapted forms in eight clinically heterogeneous mandibular reconstruction cases related to oncological treatment. These cases are presented as supplementary clinical implementations rather than direct replications of the complete methodology. Owing to differences in pathology, defect extent, reconstructive strategy, and the specific workflow stages used, the supplementary cases were summarized descriptively and were not pooled for inferential or quantitative analysis.

The first patient was a 54-year-old male whose medical data were obtained at the University Clinical Hospital in Poznań. Seven years earlier, he had undergone extensive resection of a maxillary tumor on the left side, left-sided orbital enucleation, and extended maxillectomy, followed by adjuvant radiotherapy and chemotherapy. In 2020, another surgery was performed due to ankylosis of the left temporomandibular joint. Magnetic resonance imaging (slice thickness 0.55 mm, without contrast), performed in 2023, revealed osteolysis of the mandibular head and part of the left condylar process, as well as thickening of the temporomandibular joint. Based on these findings, surgical resection of the mandibular segment and implantation of a temporomandibular joint prosthesis were planned (Figure 2A).

The second case concerned a 55-year-old male who sustained a comminuted mandibular fracture involving the left mandibular angle, the body, the mandibular tuberosity, and the condylar process as a result of a fall. During the initial surgery, all fractures were stabilized using reconstruction plates. At a follow-up visit three months later, insufficient stabilization of the condylar process was observed, with displacement of the fragment towards the mandibular midline. Consequently, a second surgical intervention was scheduled, consisting of resection of the damaged mandibular segment and implantation of a temporomandibular joint prosthesis combined with a reconstruction plate. The industrial computed tomography used for this case had a slice thickness of 0.6 mm (Figure 2B).

Supplementary Clinical Implementations

After completion of the two detailed cases, adapted variants of the workflow were applied in eight additional patients undergoing mandibular reconstruction as part of oncological treatment. The procedures involved partial mandibular resection followed by reconstruction using either a fibular graft or an iliac crest bone graft, depending on the location and extent of the defect and the selected reconstructive strategy.

The supplementary cases differed with respect to tumour location, resection extent, defect geometry, and the intended function of the patient-specific guide. Depending on the clinical requirements, the guides were used to support resection, graft preparation and positioning or transfer of the planned mandibular geometry. The workflow was therefore adapted individually for each case rather than reproduced as an identical sequence in all patients.

For each procedure, the indication, type and extent of reconstruction, workflow stages used, guide function, need for redesign, preoperative fit verification, intraoperative use, postoperative imaging availability, and guide-related complications were recorded. Not all cases included re-digitization of the modified model or FFD-based updating of the original CT-derived geometry. These cases were therefore classified as adapted supplementary implementations and were not treated as direct replications or validation of the complete closed-loop workflow. Representative examples of the patient-specific guides used in these procedures are presented in Figure 3.

2.3. Manufacturing Methodology

2.3.1. Generation and Additive Manufacturing of Preoperative Anatomical Models

The process of creating surgical models began with the analysis of computed tomography images and consultations with the medical team in order to define the requirements for the developed models. Next, a biomedical engineer carried out the image segmentation process using dedicated software (InVesalius 3.1.1). In both presented cases, identical software solutions were applied, both at the stage of segmentation and model processing, as well as during CAD design and preparation of data for additive manufacturing. For segmentation, both automatic and semi-automatic tools were used, which significantly facilitated and accelerated the creation of the patient’s three-dimensional anatomical model. Since bone tissue was the most critical structure, the threshold values were set in the range of 160–3071 in order to achieve a more detailed representation.

The created model was exported in STL format and then was subjected to further digital processing. The aim of this processing was to remove artifacts, such as undesired fragments of CT images, and to reduce the volume of the model exclusively to anatomical structures relevant for surgical planning and further design of surgical guides. The processing was carried out using specialized open-access software (GOM Inspect 2017). In the case of patient 1, the mandible was separated from the rest of the skull, which was performed with high precision using the healthy temporomandibular joint as a reference point. Ultimately, three models were created: a model of the entire skull showing the site of bone union, as well as separate models of the mandible and the maxilla. For patient 2, two models were created: the mandible and the maxilla.

The models were again saved in STL format and prepared for printing using DLP (Digital Light Processing) technology (Phrozen Sonic Mighty 8K, Phrozen Tech Co., Ltd., Hsinchu City, Taiwan) and PolyJet technology (3D Stratasys J5 MediJet, Stratasys Ltd., Eden Prairie, MN, USA), with the use of resins resistant to cracking, which allowed cutting and drilling in the models. The models for patient 2 were printed in a single process due to the large build area of the printer. In the case of DLP technology, Anycubic Basic Resin Skin (Shenzhen Anycubic Technology Co., Ltd., Shenzhen, Guangdong, China), Phrozen Aqua Resin Grey 8K (Phrozen Tech Co., Ltd., Hsinchu City, Taiwan), and Anycubic Basic Resin Light Beige (Shenzhen Anycubic Technology Co., Ltd., Shenzhen, Guangdong, China) were used, whereas for models produced with PolyJet technology, Stratasys Vero material (Stratasys Ltd., Eden Prairie, MN, USA) was applied. Key information regarding the made prints is summarized in

Table 1. Key information about the made prints.

Model	Material	Printer	Estimated Print Time	Layer Thickness
Patient 1
Face–skull	Anycubic basic resin skin	Phrozen sonic mighty 8K	13 h 37 min	0.05 mm
Mandible	Phrozen aqua resin grey 8K	Phrozen sonic mighty 4K	7 h 31 min	0.05 mm
Maxilla	Anycubic basic resin light beige	Phrozen sonic mighty 8K	13 h 46 min	0.05 mm
Patient 2
Mandible	Stratasys Vero	3D Stratasys J5 MediJet	22 h 27 min	18 μm
Maxilla

.

The post-processing of the models for patient 1 included rinsing in isopropyl alcohol (IPA), UV curing, and manual removal of supports using tools. In the case of the models prepared for patient 2, post-processing was limited only to the removal of wax supports with a water jet. The final models are shown in Figure 4.

The prepared models were delivered to the surgeons, who conducted a simulation of the planned procedures. For this purpose, they used dental wax to stabilize the models relative to each other after establishing facial symmetry. Subsequently, they defined the resection areas and removed the mandibular condyle fragments. The final stage of the simulation involved the fitting of titanium reconstruction plates to the model, which consisted of adapting them to the patient’s anatomy and fixing them to the model with single screws in order to stabilize the construction for the subsequent design process.

2.3.2. Designing of the Customized Surgical Guides

After the physical simulation, the surgeon-adjusted models were re-digitized by the engineering team. The model prepared for Patient 1 was acquired using a handheld structured-light scanner (EinScan Pro, Shining 3D, SHINING 3D Tech Co., Ltd., Hangzhou, Zhejiang, China), whereas industrial computed tomography was used for Patient 2 (GE Phoenix v|tome|x s240, GE Inspection Technologies, LP, Lewistown, PA, USA). The scanning results are presented in Figure 5A,B, respectively. As the two digitization methods were used in different clinical cases and were not assessed under a standardized metrological protocol, they were not compared quantitatively.

The re-digitized models were aligned with the original CT-derived mandibular geometry in GOM Inspect 2017 using a two-stage registration procedure. First, an initial manual alignment was performed using three corresponding anatomical landmarks identified on both geometries: the mandibular condyle, the mandibular foramen at the entrance to the mandibular canal, and the mandibular angle. The initial alignment was followed by surface-based best-fit registration using stable anatomical regions that had not been modified during the physical simulation. Regions affected by the planned resection, wax stabilization, or reconstruction-plate adaptation were not used as primary reference surfaces during the best-fit procedure. The purpose of the registration was to establish correspondence between the surgeon-adjusted physical model and the original CT-derived anatomy and to identify localized geometric differences introduced during physical planning. The alignment was assessed visually in the unchanged anatomical regions and approved by the engineering and clinical teams. No predefined numerical surface-deviation threshold was used in the original workflow.

In the two reported cases, free-form deformation (FFD) was considered and applied only as a limited supporting approach for localized geometry updating. Because the modifications introduced during physical simulation were relatively small, a comprehensive deformation procedure based on a predefined control lattice and systematic displacement of multiple control points was not performed. The role of FFD was therefore primarily analytical and methodological, supporting the interpretation and localized incorporation of differences between the re-digitized and CT-derived geometries. The aligned geometries were subsequently used as complementary references during surgical guide design. A more extensive and standardized implementation of FFD is planned for subsequent cases involving greater geometric modifications. Future applications will include formally specified registration parameters, control-lattice configuration, constrained anatomical regions, quantitative acceptance criteria, and surface-deviation analysis.

Advanced CAD software (Autodesk Inventor 2023) was used for the design of the surgical guides. The surgical guide followed a conventional patient-specific shell architecture. No structural novelty or improvement in stiffness was claimed or mechanically tested. The design task was to reproduce the contact surface of the surgeon-approved anatomy and incorporate the required cutting surfaces, reconstruction-plate recesses, fixation features, and access openings while limiting the guide volume within the surgical field. A key stage of the design process involved generating cross-sections of the mandibular segment together with the reconstruction plate, which supported case-specific fitting of the surgical guide. Cross-sections were made on the mandibular models with an offset applied. A clearance offset of 0.2 mm was applied to the guide-contact surface as an empirical fitting allowance intended to accommodate manufacturing, post-processing, and manual fitting variability. A nominal guide thickness of 1.5 mm was selected as a practical compromise between local rigidity and the need to minimize bulk in the restricted surgical field. These values were based on the team’s prior manufacturing experience and case-specific fit verification; they were not derived from a dedicated tolerance or mechanical optimization study and should not be interpreted as universally optimal parameters. Cross-sections were also created at the plate holes to determine their exact position on the mandible.

The cross-sections of surgical guide were prepared in GOM Inspect and exported in IGES format, which is a standard for CAD data exchange that allows transferring of the geometric information between different engineering systems. The cross-sections were then imported into Inventor for further design of the final geometry of the surgical guide, as shown in Figure 6A. The imported cross-sections were used to reproduce the intended contact surface of the surgical guide and to support its fit to the patient-specific bone geometry.

Based on this, a solid model of the guide was created with defined cutting surfaces, which were generated by lofting along the imported cross-sections (Figure 6B,C). The model had to meet specific clinical requirements regarding functionality and anatomical fit to the patient. Therefore, recesses for the reconstruction plate were included, as well as holes allowing for the insertion of screws to fix the plate to the bone (diameter 6.5 mm) and additional holes enabling the attachment and stabilization of the surgical guide at the implantation site (diameter 3 mm). The stage presenting the surgical guide with the recess and plate holes is shown in Figure 6D. Furthermore, the model edges were rounded to optimize its overall size, which was particularly important in the context of the limited surgical field.

The finished model of the surgical guide, saved in STL format, was printed using medical-grade resin that is resistant to sterilization and safe for patient use (

Table 2. Additive manufacturing parameters for the surgical guides.

Surgical Guide	Material	Printer	Estimated Print Time	Layer Thickness
Patient 1	NextDent surgical guide	Phrozen sonic mighty 8K	3 h 31 min	0.05 mm
Patient 2	NextDent surgical guide	Phrozen sonic mighty 8K	2 h 50 min	0.05 mm

). Before sterilization, each guide was assessed on the corresponding anatomical model together with the reconstruction plate. Verification included visual inspection of seating, manual assessment of stability, confirmation of access to the planned fixation features, and approval by the surgical team. Standardized surface-deviation maps and mechanical stability measurements were not acquired in the original clinical workflow.

2.4. Physician Evaluation Methodology

To evaluate the developed methodology and the quality of the prepared models, a short survey was conducted among the physicians involved in the process. The survey aimed to collect opinions regarding the usability and precision of the models as well as their impact on the efficiency of surgical procedures. The 12-item questionnaire was completed separately for each procedure by physicians who were directly involved in the respective surgery. The first procedure was evaluated by two physicians, whereas the second procedure was evaluated by four physicians, resulting in six completed assessments in total. The questionnaire used a five-point Likert scale, where 1 indicated a very poor rating and 5 indicated a very good rating. The survey questions included:

• The craniofacial models accurately represented the patients’ anatomical structures.
• The surgical guides were precisely matched to the patient’s anatomical structures and the reconstruction plate.
• The surface quality of the craniofacial models was satisfactory.
• The surface quality of the surgical guides was satisfactory.
• The dimensional and shape accuracy of the surgical guides was satisfactory.
• The durability of the craniofacial models was satisfactory.
• The durability of the surgical guides was satisfactory.
• The craniofacial models significantly facilitated the surgical planning process.
• The surgical guides were easy and intuitive to use during the procedure.
• The surgical guides improved the precision of the surgical procedure.
• The use of surgical guides shortened the operation time.
• The customization of surgical guides for the specific patient case positively influenced the reconstruction outcomes.

3. Results

3.1. Using of the Surgical Guides for the Maxillofacial Reconstruction

Before sterilization, the surgical guide was fitted to the mandibular model together with the reconstruction plate to confirm its seating and compatibility with the planned reconstruction (Figure 7A). Intraoperatively, following removal of the fused condylar fragment, the guide was used to support implant placement (Figure 7B). No guide-related complication was reported during the procedure. The surgeons indicated that the guide simplified the planned operative sequence and limited the need for additional intraoperative adjustments or repeated consideration of implant position, which was perceived as saving time. Postoperative computed tomography showed implant positioning considered clinically acceptable by the surgical team. As no standardized registration or geometric deviation analysis was undertaken, this assessment remained qualitative.

For Patient 2, preoperative verification showed that the guide could be seated on the anatomical model with the reconstruction plate in the intended position (Figure 8A). During surgery, occlusion and mandibular alignment were established before the guide was applied to assist with implant positioning. The procedure was completed without a reported complication related to guide use. According to the operating team, the availability of the guide reduced uncertainty during positioning and avoided additional intraoperative modifications, thereby facilitating a more efficient procedure. However, no exact reduction in operating time was prospectively recorded. Postoperative computed tomography was reviewed clinically and indicated acceptable implant placement, although quantitative positioning error was not calculated.

3.2. Physician Assessment Results

The questionnaire was completed by physicians who were directly involved in the respective procedures. Two physicians evaluated the first procedure, whereas four physicians evaluated the second procedure, resulting in six completed assessments. Owing to the small purposive sample and the unequal number of respondents between the procedures, the results were analysed descriptively using the median and range (

Table 3. Individual physician questionnaire responses and descriptive summary.

Questionnaire Item	Physician 1	Physician 2	Physician 3	Physician 4	Physician 5	Physician 6	Median	Range
Anatomical representation by models	5	5	5	5	5	4	5	4–5
Guide fit to anatomy and plate	4	4	5	5	4	4	4	4–5
Model surface quality	5	4	4	4	4	5	4	4–5
Guide surface quality	4	5	4	5	4	4	4	4–5
Guide dimensional and shape accuracy	3	4	4	4	5	4	4	3–5
Model durability	4	4	5	5	5	5	5	4–5
Guide durability	4	5	5	5	5	2	5	2–5
Support for surgical planning	2	4	5	5	5	5	5	2–5
Ease of intraoperative use	5	4	5	4	5	4	4.5	4–5
Perceived improvement in surgical precision	5	5	5	5	5	5	5	5
Perceived reduction in operating time	5	5	5	5	5	5	5	5
Perceived effect on reconstruction outcome	5	5	5	5	5	5	5	5

). Overall, the physicians provided favourable evaluations. The highest and most consistent ratings were recorded for the perceived improvement in surgical precision, reduction in operating time, and positive influence of guide customization on reconstruction outcomes; all respondents assigned the maximum score of 5 to these three items. The anatomical representation provided by the craniofacial models was also rated highly, with a median score of 5 and a range of 4–5. The lowest median score was recorded for the dimensional and shape accuracy of the surgical guides, although the median remained 4. The greatest variability was observed for the durability of the surgical guides and the extent to which the craniofacial models facilitated surgical planning, with responses ranging from 2 to 5. These results reflect the subjective assessments of the participating physicians and should be interpreted as structured clinical feedback rather than objective evidence of geometric accuracy, reduced operating time, or general clinical effectiveness.

3.3. Economic Impact of Manufacturing Surgical Guides

Table 4. Recorded workflow time and estimated direct cost for two detailed cases.

Process	Time	Cost [USD]
Patient 1
Design	12 h	238.94
3D Scanning	1 h	27.69
Printing (Total time)	42 h 30 min	23.07
Total	55 h 30 min	289.70
Patient 2
Design	7 h	139.38
Computed Tomography	15 min	476.40
Printing	25 h	227.77
Total	32 h 15 min	843.55

presents the recorded engineering and manufacturing time and the estimated direct costs associated with two detailed cases. These values describe preparation of the anatomical models and guides and should be distinguished from total hospital procedure costs. The potential procedure-level financial effect was explored using an illustrative scenario rather than a controlled health-economic analysis.

The use of anatomical models and patient-specific surgical guides was considered by the clinical team to facilitate reconstruction-plate positioning and the execution of the planned operative steps. Preoperative adaptation of the reconstruction plate and the availability of the guides reduced the need for additional intraoperative adjustments and repeated decision-making. According to the surgeons, this contributed to a more efficient operative course; however, the reduction in operating time was not prospectively measured or compared with matched procedures performed without model assistance. Therefore, no definite time-saving effect could be established.

The reference cost of mandibular reconstruction with condylar implantation, excluding the joint socket component, was approximately USD 3733.50. The costs were originally calculated in PLN and converted to USD using an exchange rate of USD 1 = PLN 3.64, corresponding approximately to the exchange rate applicable in mid-November 2025. According to the clinical team, a comparable conventional procedure performed without model assistance would require approximately 5 h. Assuming an estimated case-specific operating-time reduction of 1.5–2 h, the theoretical reduction in the total procedure cost would amount to approximately USD 560–933, corresponding to 15–25% of the reference value.

These calculations represent an illustrative scenario rather than an observed health-economic outcome. They were based on the surgical team’s estimates and the stated assumptions and were not derived from matched control cases, prospectively recorded operating-room times, hospitalization data, complication rates, or a formal cost-effectiveness analysis. Consequently, the estimated reduction should not be interpreted as evidence of confirmed cost savings or faster postoperative recovery.

4. Discussion

This study demonstrates the technical feasibility of a closed-loop digital–physical workflow for patient-specific surgical guides in two detailed mandibular reconstruction cases and documents its subsequent use in eight additional clinical implementations. In the two detailed cases, the anatomical models supported surgeon-led simulation and reconstruction-plate adaptation, the guides were verified before sterilization and used intraoperatively, and postoperative imaging provided qualitative clinical confirmation of implant placement. The supplementary cases broadened the range of clinical applications, particularly in oncological mandibular reconstruction, but remained a heterogeneous descriptive series in which adapted variants of the workflow were used.

The individual technologies incorporated into the workflow, including medical-image segmentation, virtual planning, additive manufacturing of anatomical models, CAD design, and patient-specific shell-type surgical guides, are established elements of contemporary maxillofacial reconstruction. The specific contribution of the presented approach is the physical-to-digital feedback stage. After the surgeon modifies the physical anatomical model, the adjusted geometry is re-digitized and aligned with the original CT-derived digital anatomy before the final guide is designed. This differs from a conventional unidirectional workflow, in which the guide is manufactured directly from the initial virtual plan without incorporating surgeon-introduced physical modifications. Representative examples of direct CT-based virtual surgical planning, point-of-care manufacturing, and CAD/CAM reconstruction workflows have been described previously [8,20,25], and their principal conceptual characteristics are compared with the present approach in

Table 5. Conceptual comparison of planning workflows used in patient-specific surgical guide development.

Feature	Direct CT-Based Workflow	Physical-Model Planning Without Digital Update	Proposed Closed-Loop Workflow
Starting geometry	CT-derived digital model	CT-derived digital model and printed physical model	CT-derived digital model and printed physical model
Physical surgeon-led simulation	Optional or absent	Yes	Yes
Adaptation of the reconstruction plate on the physical model	Optional	Yes	Yes
Return of surgeon-introduced physical modifications to the digital environment	No	Usually no	Yes, through re-digitization, landmark-based alignment, and surface-based best-fit registration
Role of FFD	Not applicable	Not applicable	Considered as a limited supporting approach for localized geometry updating
Final guide-design reference	Initial virtual plan	Usually the initial virtual plan or manually transferred information	Aligned CT-derived and re-digitized geometries representing the surgeon-approved physical configuration
Main potential advantage	Shorter and simpler digital workflow	Direct tactile planning and manual plate adaptation	Traceable transfer of surgeon-introduced physical modifications into the subsequent CAD process
Main additional requirement	Reliable virtual planning and direct CAD design	Manufacturing of an anatomical model	Anatomical-model manufacturing, physical simulation, re-digitization, registration, and additional CAD processing
Evidence provided in the present study	Not evaluated as a control workflow	Not evaluated as a control workflow	Technical feasibility in two detailed cases and adapted use in eight supplementary cases; no evidence of quantitative superiority

. This comparison concerns workflow structure rather than measured clinical, geometric, temporal, or economic performance, because the alternative pathways were not evaluated as matched controls in the present study.

The surgical guide itself followed a conventional patient-specific shell architecture. The study did not evaluate a new structural concept, increased stiffness, or improved mechanical stability. The design contribution consisted of reproducing the contact surface of the surgeon-approved anatomy and incorporating the required cutting surfaces, reconstruction-plate recesses, fixation features, and access openings while limiting the overall guide volume within the restricted surgical field. The empirical clearance of 0.2 mm and nominal thickness of 1.5 mm provided acceptable fit in the reported cases. However, these values were selected on the basis of the team’s previous manufacturing experience and case-specific verification and were not mechanically or statistically optimized.

The physical-to-digital feedback stage was implemented primarily through manual landmark-based alignment followed by surface-based best-fit registration of the re-digitized model with the original CT-derived mandibular geometry. Both procedures were performed in GOM Inspect 2017. The initial alignment was based on three corresponding anatomical landmarks, the mandibular condyle, the mandibular foramen at the entrance to the mandibular canal, and the mandibular angle, whereas the subsequent best-fit procedure used stable anatomical regions that had not been modified during physical simulation. This approach allowed the surgeon-adjusted physical model to be related to the original digital anatomy and enabled localized geometric differences to be identified before surgical guide design.

FFD should be interpreted as a limited supporting approach to geometry updating rather than as a fully standardized or quantitatively validated stage of the workflow. In the two detailed cases, the modifications introduced during surgeon-led physical simulation were relatively small; therefore, a comprehensive deformation procedure based on a predefined control lattice was not performed. The contribution of FFD to the final guide geometry or implant position was not quantified. Future cases involving larger physical modifications should include a formally specified FFD procedure, documented control-lattice parameters, constrained anatomical regions, and quantitative surface-deviation criteria.

The use of handheld structured-light scanning in one case and industrial computed tomography in the other illustrates that the workflow can be implemented using different digitization resources. In the first case, handheld scanning required careful acquisition and was affected by inaccessible regions and wax-covered surfaces. In the second case, industrial CT provided more complete access to external, internal, and occluded surfaces. However, no standardized metrological comparison between the two methods was performed. These observations should therefore be regarded as practical experience related to equipment availability rather than evidence of superior accuracy or reproducibility of either method.

The questionnaire results provided structured feedback from physicians who were directly involved in the procedures. The responses were generally favourable, particularly regarding the perceived improvement in surgical precision, reduction in operating time, and influence of guide customization on reconstruction outcomes. However, the questionnaire sample was small, purposively selected, and unequal between the two procedures. Moreover, the survey reflected subjective clinical impressions rather than standardized geometric, temporal, or clinical measurements. The findings therefore support usability within the participating clinical team but do not establish general clinical effectiveness.

The clinical observations should be interpreted with similar caution. In both detailed cases, the procedures were completed without reported guide-related intraoperative complications, and postoperative CT was judged by the clinical team to show clinically acceptable implant positioning. However, the study did not include standardized deviation maps, postoperative registration analysis, or calculation of implant-position error. Accordingly, these findings represent qualitative clinical confirmation rather than quantified accuracy results.

The surgical team also reported that the anatomical models and guides reduced the need for intraoperative adjustments and repeated decision-making, which was perceived as facilitating the operative sequence and saving time. Nevertheless, no prospectively defined comparator was used, and an exact operating-time reduction attributable to the workflow was not recorded. Therefore, terms such as shorter operation time or increased efficiency should be understood as case-specific clinical estimates rather than measured comparative outcomes.

The economic calculation should likewise be interpreted as an illustrative scenario. The engineering and manufacturing costs were recorded or estimated for the two detailed cases, whereas the assumed 15–25% reduction in total procedure cost was derived from an estimated operating-time difference of 1.5–2 h. The calculation was not based on matched control cases, prospectively recorded staff and equipment utilization, hospitalization duration, complication-related costs, or a formal health-economic model. It therefore illustrates a possible economic effect under the stated assumptions but does not demonstrate confirmed cost savings.

The eight supplementary cases involved clinically heterogeneous oncological mandibular reconstructions, including partial or segmental resection followed by reconstruction using a fibular graft or an iliac crest bone graft. These cases differed in tumour location, defect geometry, reconstructive strategy, guide function, and the workflow stages applied. In particular, re-digitization and FFD-based geometry updating were not used in every case. The supplementary series therefore supports the adaptability of selected workflow components across different clinical scenarios but should not be interpreted as direct replication or validation of the complete closed-loop sequence.

The principal limitations of the study are the small number of fully documented detailed cases, the retrospective and heterogeneous character of the supplementary implementations, the absence of a matched CT-based comparator, and the lack of standardized geometric deviation analysis. Guide fit was assessed visually and manually, mechanical stability was not measured, postoperative implant-position error was not calculated, and operating-time effects were based on clinician estimates. The physician survey included a small purposive sample, and the economic evaluation was scenario-based rather than formal. The study should therefore be interpreted as a clinical feasibility and workflow report.

Prospective validation is needed in a larger and more homogeneous cohort. Future studies should include standardized archiving of each digital and physical geometry, deviation analysis at successive workflow stages, postoperative implant-position assessment, prospectively recorded operating-room metrics, guide-related complications, staff and equipment utilization, hospitalization data, and health-economic indicators. These data would allow the physical-to-digital feedback step to be compared directly with conventional digital planning pathways and would clarify the clinical value of the complete closed-loop workflow.

5. Conclusions

The proposed closed-loop digital–physical workflow was technically feasible in two complex mandibular reconstruction cases involving temporomandibular joint prosthesis implantation. In both cases, the anatomical models supported surgeon-led physical simulation and reconstruction-plate adaptation, while the patient-specific surgical guides were verified before sterilization and subsequently used intraoperatively. Postoperative imaging provided qualitative clinical confirmation of implant positioning, although no standardized geometric deviation analysis was performed.

The distinguishing element of the workflow was the transfer of surgeon-introduced modifications from the physical anatomical model back into the digital environment before final guide design. In the two detailed cases, this step was implemented in a limited and localized manner because the post-simulation geometric changes were relatively small. A more extensive and standardized use of FFD is planned for future cases involving greater geometric modifications. The workflow was also applied in adapted forms in eight additional, clinically heterogeneous cases of oncological mandibular reconstruction, demonstrating the adaptability of selected workflow components rather than direct replication of the complete closed-loop methodology.

The present study does not establish quantitative superiority over direct CT-based or conventional planning workflows. Standardized geometric accuracy data, matched controls, prospectively recorded operating-time measurements, and a formal health-economic analysis were not available. The physician questionnaire reflected structured subjective feedback from a small purposive sample and should not be interpreted as objective evidence of improved accuracy, reduced operating time, lower complication risk, or better clinical outcomes. Similarly, the estimated cost reduction should be regarded as an illustrative scenario based on stated assumptions rather than a confirmed economic effect.

Overall, the findings support the technical feasibility, clinical integration, and adaptability of the workflow. Future prospective studies should include a larger and more homogeneous cohort, standardized archiving of all intermediate digital and physical geometries, a formally defined FFD procedure, quantitative assessment of guide and implant positioning errors, direct comparison with conventional planning pathways, and prospective recording of clinical, operating-room, and economic outcomes.