An integrated digital-to-surgical framework for patient-specific chest wall resection and reconstruction using 3D-printed titanium implants: proof-of-concept study

Abstract

Patient-specific skeletal reconstruction of the chest wall following extensive resections that result in large, anatomically complex defects remains challenging using conventional reconstructive techniques. We describe an end-to-end digital-to-surgical workflow framework for patient-specific chest wall resection and reconstruction using three-dimensional (3D) printed titanium implants. The four-stage workflow integrates Computed Tomography (CT) imaging data, virtual 3D anatomical modelling, virtual tumour-margin planning, computer-aided implant design (CAD), stereolithographic prototype validation, and additive manufacturing using laser-powder bed fusion (L-PBF). High-resolution CT data in Digital Imaging and Communications in Medicine (DICOM) format were converted into patient-specific 3D-rendered models of the chest wall and tumour to enable virtual delineation of tumour extent and digitally planned resection margins. These digital resection models were then used to design anatomically matched patient-specific implants. Implant designs were validated using stereolithographic prototypes before definitive manufacture in titanium alloy. The workflow underwent clinical feasibility evaluation in six consecutive patients requiring major chest wall resection and reconstruction. Across all cases, digitally planned resections and reconstructions were performed without intraoperative modification. The implants demonstrated accurate anatomical fit, secure fixation, and satisfactory restoration of chest wall anatomy. During the available follow-up period ranging from 5 months to 5 years, no implant failures or implant-related complications were observed. Two patients who underwent reconstruction for palliation subsequently died from progression of their underlying disease. This proof-of-concept single-centre study demonstrates the clinical feasibility of a digital-to-surgical workflow and provides a foundation for future multicentre evaluation.

Introduction

Surgical resection of a skeletal chest wall tumour requires a wide local excision of the tumour and full-thickness chest wall resection to ensure tumour-free margins, minimize local recurrence, and contribute to long-term survival^1,2. However, a wide local excision and full-thickness resection of the chest wall results in a large defect^3,4. Anatomical reconstruction of the resulting defect is essential to minimize thoracic deformity, restore the normal anatomical shape and structure of the chest wall, preserve its protective and respiratory functions, and when indicated, allow patients to receive adjuvant radiotherapy^4–6. Reconstruction is, however, complex and challenging, and requires a combination of pleural and skeletal reconstruction with soft tissue cover. With significant advances and refinement of surgical techniques, a variety of materials are now available to the surgeon for prosthetic chest wall reconstruction, including biologic, alloplastic, and synthetic materials (Supplementary Table 1)^6,7. However, despite the advances in surgical techniques, materials, and extensive literature available on their usage, there is little consensus amongst surgical reconstructive teams regarding the optimal material and shape of prosthetics for the skeletal reconstruction of chest wall defects⁷. Moreover, techniques, for example, synthetic meshes, methyl methacrylate composite prosthesis (MMA-CP), and titanium plates (Supplementary Table 1)^3–6,8 can restore structural integrity. Many, however, require intraoperative contouring and adaptation, resulting in variability in implant fit, operative complexity and reconstructive accuracy; hence, they have their respective limitations (Supplementary Table 1 and 2)^3–6,8.

In the past two decades, three-dimensional (3D) imaging, 3D rendering and 3D fabrication of devices have emerged as a promising and rapidly developing technology in medicine^9–11. Conventional two-dimensional (2D) modalities often have limitations in adequately demonstrating complex spatial relationships of structures. The value of 3D imaging has transformed conventional 2D imaging into accurate, manipulable 3D models that enable volumetric visualization, enhance anatomical understanding, aid decision-making, and facilitate preoperative planning^9,10,12–14.

Creating a realistic three-dimensional image of an object (3D rendering) provides an accurate representation of the anatomy and enables detailed and accurate evaluation of the body part or organ^12,15,16. The digitally rendered body part can also be virtually edited to correct defects, define resection margins, design surgical cutting guides, and joint replacement devices and implants^{11,13,17–19}. 3D printing, also known as additive manufacturing (AM), extends this capability to the fabrication of physical objects directly from virtual 3D computer-aided models using laser-powder bed fusion technology (L-PBF)^20–22. The successful application of L-PBF with titanium alloy Ti-6Al-4V^23–26, originally developed for aerospace applications^20–22 has also been translated into surgical practice, enabling the fabrication of custom-made titanium implants for precise anatomical reconstruction^{11,17,27–29}.

Applications of titanium and its alloys as biomaterials in medical, surgical, and dental devices have increased and stem from their lower elastic Young’s modulus, good fatigue performance, superior biocompatibility, resistance to infection and better corrosion resistance when compared to more conventional stainless and cobalt-based alloys (Supplementary Table 3)^30,31. These attractive properties were the driving force for the early introduction of Ti–6Al–4 V alloys as well as the more recent development of modern Ti-based alloys and orthopedic metastable β titanium alloys^30,31. Titanium is commonly used in orthopedic surgery, oral and maxillofacial surgery, dental surgery, neurosurgery and thoracic surgery as plates, screws intramedullary components, mesh, bars and rib fixation systems^31,32.

Building on these advances, we developed an integrated digital-to-surgical framework that links CT imaging, virtual tumour-margin resection planning, computer-aided implant design (CAD) and additive manufacture of patient-specific 3D-printed titanium alloy implants implementable through the Goldsmith-Bragg workflow protocol (Supplementary Note 1) for an anatomical and functional chest wall resection and reconstruction. The workflow protocol evolved through refinement of a longstanding multidisciplinary collaboration at Morriston Hospital, Swansea, between thoracic surgery and burns and plastic surgery—where traditional methods (MMA-CP) for chest wall reconstruction were used, and maxillofacial engineering services developed 3D printing technology in their laboratory^11,17. By integrating CT-imaging data in DICOM format, virtual 3D surgical planning, 3D modelling, and 3D manufacturing technology into a single workflow, the workflow framework was designed as a structured pathway for anatomically matched chest wall reconstruction (Supplementary Note 1)^11,17.

Although previous publications from our group and existing reports have described individual patient-specific implants and cases for chest wall reconstruction, there remains limited description of an integrated end-to-end workflow framework linking virtual modelling and planning of oncologic resection, computer-aided implant design, additive manufacture of patient-specific 3D-printed implants and definitive surgical implantation within a single translational workflow and clinical pathway^{17,27–29,33,34}. The principal challenge was the reliable conversion of preoperative resection planning into anatomically matched reconstruction through a clinically implementable digital workflow. The current study describes the complete end-to-end workflow, which integrates image acquisition, segmentation, virtual modelling and planning of oncological resection, computer-aided implant design, validation, additive manufacture and implantation within a defined structured pathway (Figs. 1–9). The aim of this proof-of-concept study was to evaluate the feasibility and clinical implementation of this integrated digital-to-surgical framework in six consecutive patients requiring heterogeneous, anatomically distinct, and complex chest wall resection and reconstruction. The specific aim was to determine whether preoperative resection planning and geometry could reliably inform the design and manufacture of patient-specific titanium implants for anatomically matched chest wall reconstruction.

Fig. 1. Workflow protocol overview.

Four stages indicated at the top as stage 1, stage 2, stage 3, and stage 4, and the timing for planning and performing complex chest wall reconstruction with anatomical, 3D-printed patient-specific titanium implants. Stage 1 involves assessment and planning of the surgical resection by creating a virtual 3D CT volume-rendered model of the chest wall and pathology to gain a detailed understanding of the extent of surgical resection required. Stage 2 involves the virtual surgical resection required and creating the computer-aided design (CAD) of the replacement implant. Stage 3 comprises of validation and manufacture of the replacement implant, and stage 4 comprises stepwise surgical resection and reconstruction with the anatomical replacement implant and post-surgical follow-up. This structured approach provides a clearly defined workflow for patient-specific chest wall reconstruction. (CAD = computer-aided-design; CT = computed tomography; DICOM = Digital Imaging and Communications in Medicine format; SLA Stereolithographic prototype; STL = stereolithographic files).

Fig. 9. Stage 4: Post-operative follow-up radiographic imaging.

A, B Immediate post-operative chest radiograph. C Chest radiograph at 3-month follow-up. D Post-operative follow-up CT scan at 3 months. E Follow-up CT scan at 5 years. F Scatter artifact seen on the CT scan at 3 months and resolves over time as in (E).

Results

Patient cohort and defect characteristics

The Goldsmith–Bragg workflow protocol was applied in six consecutive patients (Table 1) requiring full-thickness chest wall resection for malignant or radiation-associated pathology between 2017 and 2025. Defects involved ribs alone (n = 2); composite resections including ribs, costal cartilages, and hemi-sternum (n = 2); and sternum with adjoining costal cartilages and ribs (n = 2). Resection margins were defined preoperatively using virtual surgical planning and were reproduced intraoperatively without deviation. All resections resulted in large > 5 cm skeletal defects, which were unsuitable for primary closure, and instead required robust rigid reconstruction to restore chest wall stability, function, and anatomy. Of the six cases included in this series, Cases A (KS)²⁹, C (PM)¹⁷, D (SW)²⁷, and E (TM)²⁸ have been reported previously as individual clinical case reports describing implant design and clinical application. The present study does not re-present these cases as novel implants but incorporates them within a unified analysis of the overall workflow. The study extends these previous reports by evaluating the complete digital-to-surgical pathway across six consecutive patients, from image acquisition and virtual resection planning to implant design, validation, manufacture, and clinical implementation. The principal contribution of the present study is the integrated evaluation of this workflow together with consolidated reporting of clinical, functional and radiological outcomes across the cohort.

Table 1.

Outcomes of the Goldsmith-Bragg workflow protocol

Parameter	Ribs implant		Ribs and hemi-sternum implant		Ribs, costal cartilage and sternum implant
	Case A (KS)	Case B (ST)	Case C (PM)	Case D (SW)	Case E (TM)	Case F (AB)
Age	46	71	70	56	48	63
Gender	F	F	M	M	F	F
ASA grade	4	4	4	4	4	3
Diagnosis	Dedifferentiated chondrosarcoma	Pleomorphic liposarcoma	Chondrosarcoma Grade I	Chondrosarcoma Grade II	Metastatic breast cancer (triple negative)	Metastasizing Pleomorphic Adenoma
Bony structures resected and reconstructed	Right ribs 2-4	Left ribs 4-6	Right ribs 2-4, adjacent cartilages 2-4 and hemi-sternum	Right ribs 2-4, adjacent cartilages 2-4 and hemi-sternum	Sternum body and bilateral adjacent cartilages 2-4 and ribs 2-4	Sternum body and bilateral adjacent cartilages 2-7 and ribs 2-5
Resection intention	Curative	Curative	Curative	Palliative	Palliative	Curative
Surgery duration	6h 25 min	6h 45 min	8h 30 min	8h 15 min	8h 37 min	7h 05 min
Muscle harvest	68 min	65 min	105 min	75 min	25 min	90 min
Resection time	40 min	51 min	115 min	60 min	110 min	70 min
Pleural reconstruction	15 min	30 min	60 min	73 min	55 min	60 min
Implantation time	36 min	67 min	43 min	55 min	163 min	80 min
Flap, soft tissue cover	74 min	30 min	17 min	22 min	22 min	15 min
Total Reconstruction time + wound closure	127 min	142 min	148 min	168 min	258 min	170 min
POST-OPERATIVE
Hospital death	0	0	0	0	0	0
Complications	Nil	Nil	Excessive harvest site drainage	Excessive harvest site drainage	Nil	Nil
Length of stay	7 days	7 days	16 days	11 days	9 days	8 days
Implant related:
Infection	Nil	Nil	Nil	Nil	Nil	Nil
Movement	Nil	Nil	Nil	Nil	Nil	Nil
Dislocation	Nil	Nil	Nil	Nil	Nil	Nil
Paradoxical movement	Nil	Nil	Nil	Nil	Nil	Nil
Pain control	Epidural based analgesia	Epidural based analgesia	Epidural based analgesia	Epidural based analgesia	Epidural based analgesia	Epidural based analgesia
FOLLOW-UP at 30 days:
Death	Nil	Nil	Nil	Nil	Nil	Nil
Dyspnoea score (0-4)	0	0	0	0	0	0
Performance status (0-5)	0	0	0	0-1	0	0
Complications	Right pleural effusion drained	Nil	Nil	Nil	Nil	Nil
Implant related:	Nil	Nil	Nil	Nil	Nil	Nil
Infection	Nil	Nil	Nil	Nil	Nil	Nil
Movement	Nil	Nil	Nil	Nil	Nil	Nil
Dislocation	Nil	Nil	Nil	Nil	Nil	Nil
Paradoxical   movement	Nil	Nil	Nil	Nil	Nil	Nil
Pain	On no analgesia	On no analgesia	On no analgesia	On no analgesia	On no analgesia	On no analgesia
Aesthetic result	Yes	Yes	Yes	Yes	Yes	Yes
FOLLOW-UP at 3 months:
Death	Nil	Nil	Nil	Nil	Nil	Nil
Dyspnoea score (0-4)	0	0	0	0	0	0
Performance status (0-5)	0	0	0	0-1	0	0
Complications	Nil	Nil	Nil	Nil	Nil	Nil
Implant related:	Nil	Nil	Nil	Nil	Nil	Nil
Infection	Nil	Nil	Nil	Nil	Nil	Nil
Movement	Nil	Nil	Nil	Nil	Nil	Nil
Dislocation	Nil	Nil	Nil	Nil	Nil	Nil
Paradoxical   movement	Nil	Nil	Nil	Nil	Nil	Nil
Pain	Nil	Nil	Nil	Nil	Nil	Mild
Aesthetic result	Yes	Yes	Yes	Yes	Yes	Yes
FOLLOW-UP
Last follow-up / Status	5 years, alive, no implant related complication	5 years, alive, no implant related complication	5 years, alive, no implant related complication	9 months, died from disease progression, no implant related complication	5 months, died from disease progression, no implant related complication	6 months, alive, no implant related complication
3D Printed implant Manufacturer	Renishaw, Miskin	Renishaw, Miskin	Orthoscape, Bath	Renishaw, Miskin	Renishaw, Miskin	Orthoscape, Bath
Cost	£2,500	£4,000	£3,060	£3,960	£4,000	£5,049
Time to manufacture	10 days	10 days	14 days	14 days	21 days	28 days

Feasibility of the workflow

In all six patients, the complete digital-to-surgical workflow, from image segmentation and virtual planning through virtual surgical resection, implant manufacture, and surgical resection and reconstruction, was successfully completed without intraoperative implant modification. Virtual surgical planning enabled accurate anticipation of the shape and size of the surgical defect, and accurate implant design and positioning. No intraoperative redesign or modification of resection or implant design was required as there was no intraoperative deviation from planned resection planes. At surgery, in all six cases, complete tumour excision was achieved with clear surgical margins (Table 1).

Implant fit and intraoperative performance

In all six cases, the patient-specific 3D-printed titanium implants demonstrated precise anatomical fit to the residual ribs and/or sternum (Table 1). Fixation points aligned accurately with native bone, allowing secure anchorage with interrupted 5 Ethibond Excel^TM sutures without undue tension or need for contouring. Implants overlapped normal bone by several centimetres, providing immediate chest wall stability. No intraoperative implant instability, displacement, or interference with ventilation mechanics was observed. Successful implant placement was achieved across the six anatomically distinct chest wall defects included in this series.

Operative parameters

Duration of surgery reflected the complexity of resection and reconstruction, with longer procedures observed in composite sternum-cartilage-rib defects compared with rib-only reconstructions. The duration of surgery and type of reconstruction are described in Table 1, with individual case time detailed in Supplementary Data Cases A-F. Reconstruction time ranged between 127 min and 258 min with more complex sternal-ribs-cartilage reconstruction taking longer. Implant fixation constituted a limited proportion of total operative time, with the more complex sternal-ribs-cartilage fixation taking longer. In the remaining cases, implantation time varied within a relatively narrow range once the workflow had been established.

Early postoperative outcomes

There were no hospital deaths, no major complications, no deep surgical site infections, implant-related complications, or acute respiratory compromise attributable to the reconstruction (Table 1). All six patients achieved immediate postoperative chest wall stability with restoration of thoracic contour. There were no cases of implant movement, dislocation, paradoxical chest wall motion, or mechanical failure (Table 1). Chest drains were removed according to standard postoperative criteria, and patients progressed through physiotherapy without restriction related to the implant.

Follow-up ranged from 5 months to 5 years. Across the available follow-up period, no implant-related infection, displacement, dislocation, paradoxical chest wall motion, or mechanical failure were observed. Cases A-C remained clinically well, with Medical Research Council (MRC) dyspnoea scores of 0 and Eastern Cooperative Oncology Group (ECOG) performance status (PS) 0 at 5 years, while Case F remained complication-free, with dyspnoea scores of 0 and PS 0 at 6 months. Cases D and E, who underwent reconstruction with palliative intent, died from progression of their underlying disease at 9 months and 5 months, respectively, with no evidence of implant-related failure. At 3-month follow-up, both had MRC dyspnoea score of 0 and PS of 1.

Radiological and functional assessment

Immediate postoperative chest radiography confirmed accurate implant positioning and stable fixation (Figs. 9A, 9B). Serial chest radiographs and CT scans demonstrated maintained alignment without evidence of loosening, fracture, or displacement (Fig. 9D–F). All six patients maintained a dyspnoea score of 0–1 and reported satisfactory functional outcomes, including preservation of respiratory mechanics and tolerance of normal daily activities (Table 1).

Objective functional assessment was available in three patients prior to and at 3-months following surgery. Pulmonary function at 3-month follow-up remained comparable to baseline following reconstruction in Case A²⁹, with forced expiratory volume in 1 second (FEV₁) 92.4% prior to and 86% following surgery and forced vital capacity (FVC) 95.4% prior to and 91.9% following surgery. In Case E, there was a decrease in FEV1 from preoperative 105% to 82% following surgery and in the FVC from 108% to 75% (with no difference in the FEV1 /FVC ratio), suggesting a mild restrictive pattern of impairment that improved with physiotherapy²⁸. In Case F, cardiopulmonary exercise testing at 3 months showed preservation of exercise capacity, with postoperative values comparable to baseline (peak VO₂ 24.2 versus 21.8 mL kg⁻¹ min⁻¹ [101% versus 98% predicted], oxygen pulse 8 versus 7 mL beat⁻¹ [114% versus 102% predicted], and maximum workload 115 versus 102 W [168% versus 155% predicted]). Overall, these findings suggest preserved respiratory and functional performance following reconstruction.

Workflow implementation and protocol evaluation

Across varying chest wall sites and defect sizes, the workflow protocol was successfully implemented with satisfactory planning, implant fit, and early clinical outcomes. Quantitative geometric conformity analysis of three-dimensional deviation between the postoperative implant position and the virtually planned implant position was performed in case D (SW) as an illustrative design-verification exercise to evaluate the accuracy of the actual implant position in comparison to the preoperative plan, which was undertaken by the engineering team as a design verification step based on established experience previously described by our group²⁷. The three-month postoperative CT was segmented to generate 3D models of the implant and chest wall. These models were superimposed on the virtual preoperative plan, with registration focused on the sternum and resected ribs. Three-dimensional deviation analysis was subsequently performed in CloudCompare using mesh-distance comparison, with the virtually planned implant STL model used as the reference. Deviation values were visualised using a colour-coded surface map (Scalar field visualisation) and quantified numerically. An exploratory tolerance margin of <5 mm was used as an illustrative assessment based on expected physiological chest wall movement with respiration²⁷. Analysis demonstrated a mean surface deviation of 3.11 ± 1.40 mm (mean ± standard deviation) between the virtually planned implant and the postoperative implant position, which was considered acceptable within the context of this illustrative design-verification exercise^27,35. Hence, subsequent cases were evaluated using conventional imaging techniques. The findings support the feasibility of applying the workflow across a range of complex chest wall resection and reconstruction in appropriately selected patients and centres with access to advanced imaging and additive manufacturing infrastructure. Previously published individual cases using this workflow protocol have been referenced to provide clinical context and extended follow-up of those cases^17,27–29. These reports were produced by the same group and therefore do not constitute independent validation. Further multicentre studies are, however, required to evaluate reproducibility, scalability and generalisability.

Discussion

Our study describes the feasibility and clinical application of the patient-specific digital-to-surgical workflow for chest wall resection and reconstruction using 3D virtual surgical planning and additively manufactured patient-specific 3D titanium implants. Although individual patient-specific implants and clinical cases have been reported previously by our group^17,27–29, the present study extends these observations by incorporating additional consecutive cases and describing the complete workflow from design and manufacture to clinical implementation. The present study differs in that it defines and evaluates the complete end-to-end workflow linking CT image acquisition, segmentation, virtual oncological planning, CAD, prototype validation of the CAD, additive manufacture, and surgical reconstruction across six consecutive patients with differing anatomic defects. The principal contribution, therefore, was the description and successful implementation of an integrated workflow framework rather than a report of a novel implant design.

While certain stages of the workflow are operator dependent, particularly segmentation, virtual resection planning, CAD, and performance of surgical resection and reconstruction, all six cases were successfully completed using the same workflow pathway. The durations provided for the first two stages in the supplementary workflow protocol represent pure hands-on working time and reflect the work of one suitably qualified and trained biomedical engineer. Although operator-dependent segmentation and design steps may introduce inter- and intra-observer variability, successful implementation of the workflow was demonstrated across all six consecutive cases. A formal assessment of observer variability was beyond the scope of this study. Future investigation should evaluate reproducibility and design consistency across operators and institutions with formal geometric validation metrics using Dice coefficient, Hausdorff distance at segmentation, and inter- and intra-observer variability.

Prior biomechanical investigation and principles informed the implant design^17,27. Implant designs were derived from prior clinical experience and iterative development. The Young’s modulus of titanium alloy (113 GPa), which is substantially higher than that of cortical bone (15–20 GPa)³⁶, and its greater density (4.41 g cm⁻³ against 2.2 g cm⁻³) were considered during implant design. The cylindrical shape of the rib cage, the overall radius, and thickness of the rib were taken into consideration when designing the thickness of the rib, as were each patient’s age, size and shape of the defect, and anticipated size and weight of the respective implant for biomechanical optimisation²⁷. Nevertheless, future finite element analysis and mechanical testing may help establish and refine design criteria.

Our early results demonstrate that with our framework, anatomically matched implants can be designed, manufactured, and implanted successfully within the described framework, with accurate fit, achieving immediate chest wall stability, and favourable early clinical outcomes across a range of complex skeletal defects. Chest wall reconstruction has traditionally relied on biologic, alloplastic, and synthetic materials, which are broadly grouped as biological options, natural options, and rigid prostheses, including biological grafts, synthetic meshes, methyl methacrylate mesh composite prostheses (MMA-CP) and modular fixation systems (Supplementary Tables 1 and 4)^{5–7,37–40}. While these approaches restore chest wall function, they frequently depend on intraoperative shaping and surgeon experience; hence, they have significant limitations in reproducibility, anatomical accuracy, biological and mechanical performance, and prosthesis-related complications (Supplementary Table 2)^6,39,40. Rigid prostheses such as the MMA-CP offer intraoperative versatility; however, they are associated with complications such as infection, foreign body reaction, bone cement implantation syndrome, prosthetic dislocation leading to paradoxical movement with flail segment compromising respiration, and long-term rigidity with restrictive pulmonary deficits (Supplementary Table 2)^6,39,40. Furthermore, conventional implants (MMA-CP) are typically prepared and shaped intraoperatively, increasing surgical time and variability while reducing anatomical accuracy as preparing the implant is dependent on the operator’s individual experience and skill^3,5,6,39,40. As newer resorbable options become available, for example, synthetic and absorbable materials that consist of a β-phase Tricalcium Phosphate Ceramic (β-TCP) and Polylactic Glycolic Acid Polymer (PLGA), which induce osteo-induction and osteo-conduction for bone regeneration, limitations include partial ossification and limited long-term evaluation in load-bearing large defects³³. In contrast, the Goldsmith-Bragg workflow protocol addresses these limitations by providing patient-specific digitally planned and manufactured titanium implant tailored to each surgically created defect (Supplementary Table 5)^17,27–29. This approach enhances surgical precision with preservation of chest wall integrity and early clinical function (Table 1), thereby aligning surgical planning with anatomically matched reconstruction while reducing operative guesswork, minimizing complications, and improving consistency across cases. By integrating preoperative 3D planning, virtual modelling, and additive manufacturing, this workflow protocol offers a structured yet personalized framework that may help overcome key mechanical, biological, and logistical shortcomings of existing reconstruction strategies^{12,17,27–29}.

Several limitations should be acknowledged. This study comprises a small, single-centre case series of six patients without a control group or formal statistical analysis and does not attempt to compare clinical efficacy with a control group receiving conventional reconstruction techniques. Moreover, the small cohort size and heterogeneous pathology precluded meaningful exploration of associations between anatomical defect characteristics and clinical outcomes. The findings should be interpreted as evidence of clinical feasibility and proof-of-concept structured implementation rather than comparative effectiveness. The aim of the study was not to evaluate comparative clinical outcomes but to assess the feasibility of implementing an integrated digital-to-surgical framework across anatomically distinct and complex chest wall defects requiring robust reconstruction. The follow-up duration in the present cohort focuses on early and mid-term outcomes, and longer-term data will be required to assess implant durability, late complications, and oncological outcomes. Quality of life parameters were not available for all patients. Likewise, quantitative functional assessment was performed in three cases as a proof-of-concept evaluation, comprising full lung function in two and cardiopulmonary exercise testing in one patient. As functional outcome assessment was not performed systematically across the cohort, conclusions regarding preserved respiratory function remain preliminary. These investigations, however, demonstrated the feasibility of incorporating objective functional outcome measures within the workflow. All patients underwent routine clinical assessment of postoperative function, exercise tolerance, and chest wall stability during follow-up. The available objective data and clinical assessments were in line with satisfactory functional recovery.

Manufacturer-specific material validation data were not available for independent analysis and therefore were not included. Material-characterisation data, including residual porosity, surface roughness, fatigue performance, and lot-specific data, were undertaken by the certified manufacturers as part of their internal quality assurance practice, and these metrics were not available for independent analysis. Quantitative geometric conformity analysis was performed as a verification step in one case based on established experience; hence, it should be interpreted as illustrative^27,35. In the UK National Health Service, patient-specific titanium implants were funded through local commissioning arrangements and were not dependent on individual insurance coverage. Cost-effectiveness was not formally evaluated, and formal cost-effectiveness analysis remains necessary. While patient-specific implants incur higher upfront manufacturing costs than conventional prosthetic materials, these must be weighed against potential downstream benefits of reduced operative adjustment, shorter operating times, improved reconstruction accuracy, operative efficiency, avoidance of revision surgery, and long-term stability^{12,17,27–29,41}. In addition, implementation requires access to specialized software, engineering support, and 3D printing infrastructure, restricting widespread adoption to highly specialized centres^{12,17,27–29}. This study provides an initial feasibility framework and allows future multicenter studies to evaluate transferability, reproducibility, and wider implementation across different surgical settings. Future studies should aim to evaluate comparative effectiveness of the workflow protocol against established reconstruction methods with long-term functional outcomes and quality-of-life measures. Integration of biomechanical modelling and finite element analysis may further refine implant design and optimise load distribution across reconstructed chest wall segments⁴¹.

In summary, our study describes and demonstrates the feasibility of a structured digital-to-surgical framework in which chest wall resection geometry directly informs implant design for patient-specific chest wall reconstruction. By shifting the design and fabrication of replacement chest wall implants from intraoperative adaptation to preoperative, data-driven digital planning, this framework provides a clinically implementable pathway for reconstruction of complex chest wall defects. The workflow integrates imaging, virtual resection planning, and computer-aided design with anatomically matched manufacture of patient-specific 3D-printed titanium implants for complex chest wall resection and reconstruction, thereby aligning surgical planning with matched reconstruction and providing a foundation for future multicentre evaluation to determine the generalisability of the workflow and its potential advantages over existing reconstruction strategies.

Methods

Eligible patients were six consecutive adults requiring wide local excision of (i) three or more ribs and adjacent structures; (ii) three or more ribs and hemi-sternum including adjacent structures; (iii) sternum, cartilage and ribs and adjacent structures resulting in a large, full-thickness skeletal defect unsuitable for primary closure (Supplementary Table 6)^42,43. Cases involving a tumour affecting one rib that additionally required resection of adjacent ribs, namely one rib above and one below for wide margins, were also eligible. Inclusion criteria included adequate cardiopulmonary reserve for major thoracic surgery and availability of high-resolution thin-slice (≤1 mm) computed tomography (CT) imaging in Digital Imaging and Communications in Medicine (DICOM) format (Supplementary Table 7). Patients with extensive vertebral involvement, unresectable mediastinal invasion, or deemed unfit for general anaesthesia were excluded.

The study adhered to the principles of the Declaration of Helsinki and Good Clinical Practice guidelines. Written informed consent was obtained from all patients for surgery and the use of anonymised imaging and intraoperative data. Under guidance from the NHS Research Ethics Centre, as the interventions were undertaken as an innovative procedure for individual patient benefit, approval was obtained from the Swansea Bay University Health Board Clinical Governance Department, and the project was registered locally, with no reference number assigned.

The workflow protocol followed four principal stages (Fig. 1), namely, (i) Preoperative image acquisition, segmentation, and assessment. With DICOM CT imaging data and segmentation using Mimics Medical 20.0 software (Materialise NV, Technologielaan 15, 3001 Leuven, Belgium), the generation of patient-specific 3D virtual anatomical models and a detailed 3D-volume-rendered model of the tumour and chest wall provided insights into the tumours geometric size, volume and relationship to surrounding thoracic structures, which facilitated the surgical reconstructive team to plan the extent of surgical resection required^17,27–29. (ii) Virtual surgical planning and computer-aided design (CAD) allowed digital editing of segmented data of the chest wall and tumour using Geomagic Freeform Plus software (3D Systems Inc., Rock Hill South Carolina, USA and presently Hexagon AB (Publ), Lilla Bantorget 15SE-111 23 Stockholm, Sweden) to simulate tumour excision, define resection margins, and design the patient-specific titanium implant that mimicked the anatomical shape, size and contour of the resected area^17,27–29. (iii) Validation and manufacture. First, a virtual validation was performed, then a physical validation of the design using stereolithographic (SLA) prototypes, followed by definitive manufacturing of the titanium implant by laser-powder bed fusion (L-PBF), and sterilization of the final product (implant)^17,27,28. (iv) Intraoperative application was carried out with stepwise surgical assessment, surgical resection, and chest wall reconstruction with the anatomical replacement 3D-printed patient-specific titanium implant (Fig. 1)^17,27,28, with the timeline described in Table 2 and a detailed workflow protocol described in Supplementary Note 1.

Table 2.

Timings for the Goldsmith-Bragg workflow protocol

Stage	Step summary	Estimation duration	Comments
Stage 1. Preoperative image acquisition and segmentation	High-resolution CT, Segmentation of bones and tumour, STL generation	2–3 h	Ensure ≤1 mm slice thickness
Stage 2. Planning virtual surgical resection and designing implant	Import into Geomagic Freeform Plus, define margins, simulate resection	1–1.5 h	Surgeon–engineer joint review recommended
	Create implant design, add fixation features, design cutting guides	4–6 h	Surgeons’ confirmation for oncologic margins and design
Stage 3. Validation and 3D printing	Validation with 3D-rendered design and with prototypes 3D printed in laboratory	8–10 h	Overnight print; surgeon verification next day
	Manufacture titanium implant (DMLS/SLM). Finishing, polishing, passivation	12–16 h	Orthoscape 3D Metal Printing Ltd. Bath BA1 1UD
	Decontamination, sterilization and packing	~12 h	Surgeon verification
		Total stage 3: 10–28 days
Stage 4a. Operative resection and reconstruction	Anaesthesia, epidural siting, positioning	45-90 min	Lateral or supine position depending on tumour site
	Muscle harvest (typically latissimus dorsi)	60-100 min	Typically, muscle harvest followed by resection
	Tumour resection, implant fixation, flap soft tissue cover	165-350 min	En bloc excision with clear margins
Stage 4b. Post-operative aftercare	Recovery, HDU, ward, chest physiotherapy, pain, and drain management	Day 0-7	Excessive harvest site drainage may delay discharge
Stage 4c. Follow-up care	Clinical review and imaging	4-6 weeks; follow-up to 5 years	Assess wounds, implant position, stability and chest wall dynamics

Stage 1: Image acquisition, segmentation and assessment

Thin-slice (≤1 mm) CT scans of the thorax were acquired in DICOM format with the patient in a neutral supine position to minimize respiratory motion artefacts (Figs. 1 and 2A, 2Aa and Supplementary Table 7)^{15–17,27,44,45}.

Fig. 2. Stage 1: Preoperative image acquisition and segmentation.

A, Aa Computed tomography (CT) scan of the thorax and abdomen: Obtained with the patient in a supine position, arms above the head, a 0-gantry tilt and 1.0 mm slice thickness, and imported into Mimics Medical 20.0 software (Materialise, Leuven, Belgium) software. B, Bb A 3D virtual model of the patient’s chest wall with the tumour: Created with the ‘Thresholding tool’ in Mimics Medical 20.0 software and ‘Region-Grow’ applied to isolate bone, cartilage and tumour from soft tissues. The ‘Edit Mask’ aids removal of unwanted parts. The segmentation is refined manually to remove surrounding soft tissue and ensure accuracy of tumour-rib interface. The 3D STL file of the skeletal and tumour structures is then imported into Geomagic Freeform Plus 3D computer-aided design (CAD) software. C, Cc A 2 cm positive offset of the tumour in the soft tissue planes: This is created in Geomagic Freeform Plus to define all around 2 cm clear soft tissue margins, and 4–5 cm clear bone resection margins along the length of the ribs/sternum.

CT data were imported into Mimics Medical 20.0 (Materialise, Leuven, Belgium) for segmentation and generation of the computer-aided 3D-rendered model of the chest wall and tumour^{17,27,46–48}. Manual bone threshold segmentation was preferred to automatic segmentation, which allowed precise differentiation between tumour, ribs, sternum, and costal cartilages^17,27,28. The segmented data were reconstructed into 3D stereolithographic (STL) files representing the tumour and skeletal structures in depth and detail.

Manual adjustment by visual verification ensured anatomical fidelity (Fig. 2B, 2Bb)^46–48, particularly at tumour–bone interfaces and rib curvatures^11,17,27. The segmentation defined the spatial relationship of the tumour to adjacent ribs, costal cartilages, sternum and underlying pleura and thoracic viscera, and served as the foundation for planning the virtual resection.

For virtual planning, STL files of the ribs, sternum, and tumour were imported into Geomagic Freeform Plus 3D computer-aided design (CAD) software (3D Systems, Rock Hill, United States)^17,27,28. Using a haptic device interface, resection margins were digitally planned, typically extending 2 cm beyond tumour boundaries in all soft-tissue planes (Fig. 2C, 2Cc)^{1,17,27,42,43,49}. On bone, depending on the grade of the tumour and presence or absence of an intervening articulating joint, 2–5 cm margins along ribs were deemed required and were planned to ensure clear microscopic margins (Figs. 1 and 2)^17,27,28,50. To do so, the tumour was digitally “grown” by 2 cm to represent oncological margins for the soft tissues (Fig. 2C, 2Cc)^17,27–29. For the rib(s), 4–5 cm margins were adopted along the length of the involved rib(s) in its medial and lateral extent [Supplementary Note 1].

Anatomical modelling was performed using a haptic-enabled workstation for organic modelling. Although equivalent modelling can be undertaken using conventional mouse-based interaction, the haptic interface provided enhanced three-dimensional manipulation and tactile feedback during segmentation and contour refinement with the Freeform software. No formal comparison between interfaces was, however, undertaken.

Stage 2: Planning the virtual resection and implant design

For virtual simulation of resection, digital cutting lines were mapped across involved ribs and the sternum, ensuring adequate clearance superiorly and inferiorly (including one normal rib above and below the tumour level) (Figures and 3A, 3Aa). This virtual plan allowed visual simulation of the resection and provided preoperative insight into the expected defect size and geometry (Fig. 3B, 3Bb)^17,27–29. The resected area served as the template model for the patient-specific implant design required for reconstructing the defect (Fig. 1)^17,27–29.

Fig. 3. Stage 2: Planning virtual surgical resection and designing the implant.

A, Aa Tumour, ribs and sternum in 3D, and digital resection planes along the anatomical boundaries of the ribs and sternum with delineated and precisely defined tumour margins: Visualized using the haptic interface in Geomagic Freeform Plus. B, Bb Simulated virtual resection of the ribs and sternum: This is achieved using the ‘Trim’ or ‘Boolean Subtract’ function to digitally excise the tumour-bearing region and simulate the intended surgical resection defect. The resected model serves as a template for patient-specific implant design. C, Cc Creating a conformal 3D anatomical implant: The 3D implant is fashioned to mimic the shape and contours of the resected ribs, cartilage, and sternum.

To design the implant, following virtual resection, the resected chest wall segment was emulated to create a conformal implant, fashioned to mimic the shape and contours of the ribs, cartilage and sternum and thus replicate the natural anatomy (Fig. 3C, 3Cc)^17,27–29. Using Geomagic Freeform Plus, the implant was designed as a single titanium component that restored the curvature and contour of the chest wall and faithfully reproduced the resected ribs and hemi-sternum/sternum (Fig. 1, Fig. 3C, 3Cc)^17,27–29.

The implant design incorporated (Fig. 4A–4C)^17,27:

• Fixation holes for anchorage to the remaining ribs or sternum (Fig. 4A, 4B).

• Overlapping rabbeted edges for the implant to sit on the bone stumps and slot in (Fig. 4C).

• Perforations to permit drainage and facilitate tissue integration (Fig. 4B).

• Anatomical curvature to maintain thoracic shape and optimize respiratory dynamics (Figs. 4A, 4C).

Fig. 4. Stage 2: Computer-aided designing of implant.

A Computer-aided designed, 3D rendered anatomical implant: Thickness of the implant for the rib section is kept at 4–5 mm, and 6–7 mm for the sternum section (depending on the anatomy of the patient). Fillets are added at the junction of the rib arms with the sternum, to avoid stress concentration areas during flexion and lateral bending movements of the patient. B Fixation holes: Each 2.5 mm in diameter, are fashioned in pairs around the contour of the sternum with minimum edge-to-edge distance of 3 mm between the holes, and in groups of three in each rib with a minimum edge-to-edge distance of 1.5 mm. These fixation holes will be used to attach the implant to the bone with interrupted size 5 Ethibond Excel Polyester sutures (Ethicon, Somerville, USA) at surgery. Perforations (lattice) each of 0.5 mm diameter are added to the main body of the sternum section to reduce weight, permit tissue ingrowth and allow fluid drainage. C Rabbet (stepped) edges: The body of the implant (sternum and rib sections) is designed to be thicker (sternum 6–7 mm and ribs 4–5 mm) than their overlapping area of attachment (rabbet thickness 1.5 mm) at the implant-bone interface. This is to allow the implant to sit on top of the healthy bone stumps and slot into the bony defect, thereby providing secure fixation and preventing future dislocation and paradoxical movement.

Considering that the Young’s modulus of titanium alloy (113 GPa) is substantially higher than the cortical bone (15-20 GPa), the implant thickness was kept at 4–5 mm for the rib section and 6–7 mm for the sternum section to achieve a robust reconstruction with a profile that was lower than the resected anatomy^27,36. Fixation holes of 2.5 mm diameter were added in pairs around the contour of the sternum and in groups of three in each rib. In the sternal component, adjacent holes were separated by a minimum edge-to-edge distance of 3 mm (Fig. 4B), and in the rib components, adjacent holes were separated by a minimum edge-to-edge distance of 1.5 mm according to standard manufacturing specifications^17,27. Perforations of 0.5 mm diameter were manually added to the sternum section to reduce weight; permit tissue ingrowth and facilitate soft tissue integration; allow fluid drainage to prevent fluid from collecting behind the implant and predispose to infection; and also to provide soft tissue fixation points if required at surgery^17,27. Design parameters including implant thickness (sternum thickness 6–7 mm and rib thickness 4–5 mm), perforation geometry and distribution, overlap length (3–5 cm) and rabbet geometry (rabbet thickness 1.5 mm) were conceptualised and tested over six cases^{17,27–29,36,51}. Hence, these design parameters should be regarded as empirically derived and applied every time in the present six cases. The parameters were considered satisfactory and did not require modifications over subsequent cases. Furthermore, geometric parameters were based on previous clinical experience and iterative design refinement rather than automated software generation^17,27.

The durations given for stage 1 and stage 2 (Supplementary Note 1), moreover, represent pure hands-on working time and reflect the work of one suitably qualified and trained biomedical engineer.

The intended bone resections required a high level of precision to accommodate the custom prosthesis, as the error in bone resection magnifies with an increased number of resection planes around the tumour⁵². 3D-printed patient-specific cutting guides were designed using Geomagic Freeform Plus to provide a cutting platform that confined the saw blade to follow a pre-determined path and replicate the surgical plans accurately at surgery.

Stage 3: Validation and manufacturing

For virtual validation, all design parameters, including the intended bone resection and 3D-rendered CAD were reviewed jointly by the surgeon and the biomedical engineer to ensure anatomical and surgical feasibility (Fig. 1)^17,27–29. A virtual fit validation with the surgical reconstructive team was performed by superimposing the implant onto the model of the chest wall defect (Fig. 5A, 5Aa), and thereby confirming geometric conformity, alignment of all edges to the underlying defect and accuracy of fixation points.

Fig. 5. Stage 3: Validation and 3D printing.

A, Aa Virtual fit validation of the 3D implant: The implant image is superimposed on the model of the chest wall defect, which helps to confirm the fit, geometric conformity, alignment of all edges to the underlying defects, and accuracy of fixation points. This fit validation is performed jointly with the surgical reconstructive team. B, Bb Physical fit validation: This is performed with a full-scale (1:1) sterilisable stereolithographic resin prototype (SLA) of the 3D test model of the implant and the resection defect. This physical fit validation is performed jointly with the surgical reconstructive team. C, Cc The final 3D laser powder-bed fusion-printed titanium implant.

The proposed implant and resection model stereolithographic prototype (SLA) were then printed using photopolymer resin at 1:1 scale for preoperative validation with 3D-printed prototypes (Figs. 1 and 5B). These models were examined by the surgical team to verify anatomical accuracy, confirm resection lines, and assess implant–host fit (Figs. 1 and 5Bb)^17,27. Any discrepancies were digitally corrected before final fabrication.

For titanium fabrication, the final STL file was transferred to an L-PBF manufacturer (e.g., Orthoscape 3D Metal Printing Ltd. Bath, United Kingdom, BA1 1UD; and Renishaw, Metal 3D printing, Miskin, United Kingdom, CF72 8XY). The manufacturers and our maxillofacial laboratory comply with ISO 13485-certified quality systems and UK MDR requirements for custom-made medical devices. The definitive implant was manufactured from medical-grade titanium alloy (Ti-6Al-4V ELI Grade 23) using laser-powder bed fusion (L-PBF), with direct metal laser sintering (DMLS) or selective laser melting (SLM) processes according to the manufacturing facility used (Fig. 5C, 5Cc)^{11,17,27–29,53}. The implants were manufactured and processed under the manufacturers’ certified quality-management systems, and independent material-characterisation data, including residual porosity, surface roughness and fatigue performance, were not available for analysis in the present study^17,27–29.

When printing was complete, an ultrasonic wand was used to ensure the residual powder was removed. The part was heat-treated in a vacuum furnace running to a peak temperature of 850°C. Then, it was removed from the plate using a band saw, cutting through the hollow supports. Any remaining support structures were removed using side cutters. The part underwent bead blasting and a three-stage tumble deburring. All holes were measured and re-drilled to ensure dimensional accuracy. Then various grit polishing wheels were used for surface finishing. The part was steam cleaned to remove any polishing residue and put through nitric acid passivation. As a result of the ultrasonic tanks used in the passivation process, any powder left in the part after ultrasonic wand either sintered to the walls or was removed during heat treatment.

On receipt of the implant from the manufacturer, dimensional verification of the produced implant was limited to measuring key functional dimensions with Vernier callipers and comparing them with corresponding CAD-derived dimensions prior to performing the essential final physical fit confirmation on the anatomical 3D-printed stereolithographic anatomical prototype model of the healthy rib and/or sternum (Fig. 5). A formal dimensional acceptance threshold and quantitative deviation dataset were not recorded and were therefore unavailable for analysis. For confirmation of implant model fit, following decontamination, the final implant was tested on the previously 3D printed full-scale resin prototype (SLA) of the resection defect. The anatomical fit was physically performed, and alignment of the fixating holes was confirmed with the surgical team before sterilization according to the manufacturing company specifications (Supplementary Note 1).

The implants were steam sterilised in-house according to the strict manufacturer’s recommendations (Supplementary Note 1).

Stage 4: Surgical resection and reconstruction

Under general anesthesia, harvest of the muscle flap (commonly the latissimus dorsi muscle) as a pedicle flap was typically performed first, verifying the flap’s reached the anticipated defect to cover the entire area using a sterile tape measure (Fig. 6A–6D)^17,27–29.

Fig. 6. Stage 4: Muscle flap harvest.

A Latissimus dorsi muscle harvest: This is commenced under general anaesthesia, with a double-lumen endobronchial tube for ventilation, and the patient secured on the operating table with bolsters or vacuum bean bag, and pressure points padded to avoid nerve injury. B, C The harvested latissimus dorsi muscle: Detached from its inferior attachment on the iliac crest and thoracolumbar fascia posteriorly. D Primary closure of the donor wound. The harvested muscle is placed in the axilla, and the donor wound is closed over 14 F redivac drain(s). E Pre-marked lines of incision: The incision commences under general anaesthesia, with the patient in a supine position and arms spread out on each side. F, G En Bloc resection of the soft tissues and tumour with 2 cm clear all around macroscopic margins. H The resected tumour and chest wall. The specimen is marked with sutures to identify orientations to structures.

The planned en bloc resection was performed under general anesthesia with single-lung ventilation. An epidural catheter was inserted for post-operative pain relief and to facilitate physiotherapy. The tumour-bearing ribs, sternum, and adjacent pleura were removed en bloc with pre-defined margins (Fig. 6E–6H), guided by the 3D prototype, surgical guide(s), and replacement implant (Fig. 7A–7C)^17,27–29. Intraoperative frozen section analysis was not deemed necessary⁵⁴.

Fig. 7. Stage 4: Precise surgical resection and fit validation.

A The surgically created resection defect. B, C The 3D-printed anatomical Titanium implant and cutting guide help to achieve precise, anatomical bone resection. C Physical validation of fit with resin SLA prototype.

Prior to reconstructing the defect (Figs. 7C, 8A), the 3D test model of the implant was used to first test that the implant would sit on the bony stumps of the resected area and slot into the surgical defect perfectly (Fig. 7C).

Fig. 8. Stage 4: Surgical reconstruction of chest wall defect.

A Surgically created chest wall defect for reconstruction. (Aa) Pleural reconstruction with GORE^® DUALMESH^® Biomaterial mesh. The mesh is secured with 2-0 Ethibond Excel^TM interrupted sutures to the under surface of the ribs and sternum. B, Bb Anatomical, skeletal reconstruction. The 3D implant is secured with size 5 Ethibond Excel^TM interrupted sutures. C Soft-tissue reconstruction with muscle flap. Cc Final result following skin closure.

For pleural reconstruction, an underlying mesh (GORE^® DUALMESH^® Biomaterial) was first applied to the defect to recreate the pleural barrier with interrupted 2-0 Ethibond Excel^TM sutures, which were tied to secure the mesh once the implant sutures were placed but not tied (Fig. 8A, 8Aa)^17,27–29.

To achieve implant fixation, the custom titanium implant was positioned in the defect to restore the contour and rigidity of the resected chest wall. Fixation was achieved using interrupted 5 Ethibond Excel^TM sutures through the pre-designed holes, ensuring durable stabilization yet allowing for flexibility of movement during breathing (Fig. 8B, 8Bb)^17,27–29. The implant overlapped normal ribs and sternum by at least 3–5 cm on either side to prevent displacement into the defect and paradoxical movement at respiration^17,27–29.

Finally, for soft-tissue reconstruction, the implant was then covered with a well-vascularized muscle flap, typically latissimus dorsi (and depending on the location, pectoralis major or rectus abdominis were also available), to provide soft tissue coverage, and reduce infection risk and improved cosmesis (Fig. 8C, 8Cc)^{12,17,27–29}. When local tissues were damaged or scarred, free tissue transfer was a viable option with careful flap design orientated to recipient vessels in the axilla.

Postoperative management in a high-dependency care unit included management of chest drainage, pain relief with thoracic epidural analgesia, early physiotherapy, and respiratory support as required^17,27–29. At follow-up, patients were monitored for wound healing, implant stability, and pulmonary function. Dyspnoea was graded using the Medical Research Council (MRC) breathlessness scale (0–4) and functional performance using the Eastern Cooperative Oncology Group (ECOG) scale (0–5). Short and long-term follow-up chest radiograph (Fig. 9A–9C) and CT imaging (Fig. 9D–9F) were performed to assess for proper implant position, integration, and surveillance for tumour recurrence^17,27–29.

Author contributions

IG was responsible for conceptualization, method development, data collection, writing the original draft, and preparing the figures based on his initial study, which was published in 3D Printing in Medicine. IG performed the literature search with assistance from RW, a Swansea Bay University Health Board librarian. TB and IG obtained permission from the Clinical Governance team of the Health Board for the study. All authors were responsible for reviewing and editing the manuscript.

Data availability

The tested technology was self-designed and not purchased, borrowed, or donated. All figures are original and were created by the authors. The author has full control of the design of the study materials, methods used, outcome parameters, and production of the written manuscript. The dataset generated and analysed is not publicly available due to patient confidentiality and institutional governance requirements but is available from the corresponding author upon request and subject to approval by the Swansea Bay University Health Board.

Declarations

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s44385-026-00113-6.