Patient-Specific 3D Printed Soft Models for Liver Surgical Planning and Hands-On Training

Background: Pre-surgical simulation-based training with three-dimensional (3D) models has been intensively developed in complex surgeries in recent years. This is also the case in liver surgery, although with fewer reported examples. The simulation-based training with 3D models represents an alternative to current surgical simulation methods based on animal or ex vivo models or virtual reality (VR), showing reported advantages, which makes the development of realistic 3D-printed models an option. This work presents an innovative, low-cost approach for producing patient-specific 3D anatomical models for hands-on simulation and training. Methods: The article reports three paediatric cases presenting complex liver tumours that were transferred to a major paediatric referral centre for treatment: hepatoblastoma, hepatic hamartoma and biliary tract rhabdomyosarcoma. The complete process of the additively manufactured liver tumour simulators is described, and the different steps for the correct development of each case are explained: (1) medical image acquisition; (2) segmentation; (3) 3D printing; (4) quality control/validation; and (5) cost. A digital workflow for liver cancer surgical planning is proposed. Results: Three hepatic surgeries were planned, with 3D simulators built using 3D printing and silicone moulding techniques. The 3D physical models showed highly accurate replications of the actual condition. Additionally, they proved to be more cost-effective in comparison with other models. Conclusions: It is demonstrated that it is possible to manufacture accurate and cost-effective 3D-printed soft surgical planning simulators for treating liver cancer. The 3D models allowed for proper pre-surgical planning and simulation training in the three cases reported, making it a valuable aid for surgeons.

Gels 2023, 9,339 3 of 15 of the 3D model and lack of quantitative quality validation in the production process.
(1) Souzaki et al. [22] and Zein et al. [23] manufactured liver cases using the material jetting technology, which is very expensive and limits its use in hospitals. (2) Witowski et al. [24] reported a more cost-effective approach but presented challenges in the anatomical accuracy of the final model. (3) Tan et al. [4] developed a liver case taken from an online database for surgical simulation training and, therefore, the case was probably not used for preoperative surgical planning. Finally, (4) in most of the published cases, there is no quantitative validation of the models [21][22][23]. However, Tejo-Otero et al. [21] were not only able to obtain a 3D physical model but also represent the mechanics of the liver through the use of different hydrogels and silicones.
Amongst the different materials that have been used for the present application, hydrogels and silicones appear as the most common soft materials. On the one hand, hydrogels are soft materials that are mainly used in tissue engineering applications, although they have been used in some prototypes [16,21]. Nevertheless, they have some disadvantages: (1) a lot of preparation and processes are necessary; for example, Forte et al. had to perform one or two freeze-thaw cycles; (2) they are not consistent enough, posing challenges in the manufacturing and repeatability, as reported by Tejo-Otero et al. [21]; and (3) they might degrade very quickly compared to silicone-made prototypes. One of these examples is cellulose, commonly used in certain bio-applications, yet not the best for the present study for the reasons mentioned above [26,27]. On the other hand, silicones are a synthetic polymer made up of silicon, oxygen, carbon and hydrogen [28]. They are very versatile materials that can be formulated into various types of materials, such as elastomers, gels (when it is in a semi-solid state) or adhesives, showing the possibility of being used in a wide range of applications [28]. They are of interest in the biomedical field due to different factors [28] depending on the target use: (1) they are firm and flexible; (2) stable to temperature and chemical conditions; and (3) they are inert and non-toxic. In this way, they are widely used in the biomedical field in a wide range of applications, such as breast implants [29], prostheses [30], hypertrophic burn scars [31] or phantoms [32]. The latter application can be performed with hydrogels, although, as mentioned before, they are not very consistent, and their lifespan is very short, not enough to be used for purposes apart from surgical planning, such as patient education [33] or for medical school interns.
Therefore, the aim of the present article is to report the full process of three additivemanufactured liver models as an aid in the surgical planning of three complex paediatric liver cases by explaining the complete workflow for the development of each case, namely: (1) medical imaging acquisition; (2) segmentation; (3) 3D printing; (4) quality control/validation; and (5) cost. Figure 1 shows the process of the present research study. 3D model and lack of quantitative quality validation in the production process. (1) Souzaki et al. [22] and Zein et al. [23] manufactured liver cases using the material jetting technology, which is very expensive and limits its use in hospitals. (2) Witowski et al. [24] reported a more cost-effective approach but presented challenges in the anatomical accuracy of the final model. (3) Tan et al. [4] developed a liver case taken from an online database for surgical simulation training and, therefore, the case was probably not used for preoperative surgical planning. Finally, (4) in most of the published cases, there is no quantitative validation of the models [21][22][23]. However, Tejo-Otero et al. [21] were not only able to obtain a 3D physical model but also represent the mechanics of the liver through the use of different hydrogels and silicones.
Amongst the different materials that have been used for the present application, hydrogels and silicones appear as the most common soft materials. On the one hand, hydrogels are soft materials that are mainly used in tissue engineering applications, although they have been used in some prototypes [16,21]. Nevertheless, they have some disadvantages: (1) a lot of preparation and processes are necessary; for example, Forte et al. had to perform one or two freeze-thaw cycles; (2) they are not consistent enough, posing challenges in the manufacturing and repeatability, as reported by Tejo-Otero et al. [21]; and (3) they might degrade very quickly compared to silicone-made prototypes. One of these examples is cellulose, commonly used in certain bio-applications, yet not the best for the present study for the reasons mentioned above [26,27]. On the other hand, silicones are a synthetic polymer made up of silicon, oxygen, carbon and hydrogen [28]. They are very versatile materials that can be formulated into various types of materials, such as elastomers, gels (when it is in a semi-solid state) or adhesives, showing the possibility of being used in a wide range of applications [28]. They are of interest in the biomedical field due to different factors [28] depending on the target use: (1) they are firm and flexible; (2) stable to temperature and chemical conditions; and (3) they are inert and non-toxic. In this way, they are widely used in the biomedical field in a wide range of applications, such as breast implants [29], prostheses [30], hypertrophic burn scars [31] or phantoms [32]. The latter application can be performed with hydrogels, although, as mentioned before, they are not very consistent, and their lifespan is very short, not enough to be used for purposes apart from surgical planning, such as patient education [33] or for medical school interns.
Therefore, the aim of the present article is to report the full process of three additivemanufactured liver models as an aid in the surgical planning of three complex paediatric liver cases by explaining the complete workflow for the development of each case, namely: (1) medical imaging acquisition; (2) segmentation; (3) 3D printing; (4) quality control/validation; and (5) cost. Figure 1 shows the process of the present research study.

Results
The 3D physical models were manufactured before the planned surgery, so the surgical team was able to prepare the case, simulate and practice using the models in advance. The prototypes were a 1:1 scale of the patient's organs and, consequently, gave an impression of what to expect during surgery. This can be confirmed by the comparison of the tumour removed from case #2 (see Figure 2D). Additionally, the soft consistency of the prototype permitted the use of the surgical instruments that would be used in performing the operation. In summary, the models gave surgeons a new tool for surgical planning and pre-surgical simulation training.

Results
The 3D physical models were manufactured before the planned surgery, so th gical team was able to prepare the case, simulate and practice using the models in adv The prototypes were a 1:1 scale of the patient's organs and, consequently, gave an im sion of what to expect during surgery. This can be confirmed by the comparison tumour removed from case #2 (see Figure 2D). Additionally, the soft consistency prototype permitted the use of the surgical instruments that would be used in perfo the operation. In summary, the models gave surgeons a new tool for surgical pla and pre-surgical simulation training.   Figure 3 shows the surgical planning prototypes of the hepatobiliary oncological cases. The silicone used gives the desired transparency that surgeons are looking for, which is very difficult to achieve with other types of hydrogels or silicones. The transparency gives surgeons the advantage of observing the exact position of internal blood vessels and their relation to the tumour and anatomy. Moreover, as soft silicone was used, it was possible to practice with it by using medical surgical tools such as lancets or Kelly forceps with surgical sutures.

Validation
There is no significant difference between the original organs and t Figure 2). Regarding Cases #1 and #2, the biggest error was measured in located outside the phantoms. This happened because it was not poss parts. Therefore, these areas appear in red, which indicates a higher er marises the parameters obtained from the validation of the three cases mensional error is low, less than 3.35 mm, 4.74 mm and 2.1 mm for each Apart from a CT validation, the usefulness and accuracy of the 3DP rep cally validated by comparing them at the time of surgery with the remov Figure 2D). These prototypes overcame the drawbacks of the previou reasons: (1) the use of transparent silicone, using highly accurate c thanks to the defined production process and the 3D technologies prov feedback of the hepatobiliary anatomy; (2) this prototype was used for la ies during the operation thanks to the use of sterilisable materials; and appreciated the liver softness in surgical planning, compared to other

Validation
There is no significant difference between the original organs and the phantoms (see Figure 2). Regarding Cases #1 and #2, the biggest error was measured in the blood vessels located outside the phantoms. This happened because it was not possible to scan those parts. Therefore, these areas appear in red, which indicates a higher error. Table 1 summarises the parameters obtained from the validation of the three cases. All in all, the dimensional error is low, less than 3.35 mm, 4.74 mm and 2.1 mm for each case, respectively. Apart from a CT validation, the usefulness and accuracy of the 3DP replicas can be clinically validated by comparing them at the time of surgery with the removed specimen (see Figure 2D). These prototypes overcame the drawbacks of the previous ones for several reasons: (1) the use of transparent silicone, using highly accurate casting techniques thanks to the defined production process and the 3D technologies providing good visual feedback of the hepatobiliary anatomy; (2) this prototype was used for last minute enquiries during the operation thanks to the use of sterilisable materials; and (3) the surgeons appreciated the liver softness in surgical planning, compared to other previously used prototypes.  Table 2 summarises the cost in terms of materials and labour. The labour cost is the combination of engineers for the segmentation and 3D printing, as well as the necessary post-processing. Regarding oncologists and radiologists, time is not taken into consideration since their work is something commonly performed in operations. It can be highlighted that this price is lower than that of prototypes manufactured using other techniques, such as material jetting, which may cost 2000 euros (€).  Table 3 shows the different times needed for each part of the process. The 3D printing parts in the present table show the time of the 3D model, although it must be taken into consideration that most of these parts are 3D printed with other works needed for other applications or clients. In the SLS part, post-processing is also considered.

Discussion
This study demonstrates that it is possible to manufacture 3D-printed soft surgical planning prototypes for a better simulation experience. For the present research, three different cases were taken into consideration, in which the full production process was included: from the DICOM acquisition to their manufacture, as well as a final validation and summary of costs.
The silicones used, as mentioned, are softer than the common materials used in 3D printing techniques, such as PLA or PA12. One of the silicones used had a value of 38 Shore A, which is a soft value, although it is not as soft as the liver tissue. According to different studies, the measured liver tissue is in the range of Shore 00 [7,10,11]. Although in this sense, it is possible to achieve 100% mechanic mimicking, which is the best option when hydrogels are not used. Additionally, silicones, like the ones used, offer the possibility of seeing the inner anatomical structures, which are not possible to be seen during the operation as well as with other types of materials.
Aside from this enhanced soft texture simulation experience, it is possible to use these models to improve patients' experience before an operation, thanks to the better understanding they manifest when viewing and touching the 3DP replicas, as opposed to only biplanar CT images. Additionally, as an indirect 3D printing technique was used, it is possible to reuse these moulds for manufacturing more prototypes in the future for further teaching purposes.
The results of the validation of the three cases are in concordance with Bücking et al. [34], which measured an error of 1.3% between the phantom model and the patient's liver. Moreover, Witowski et al. [35] measured an average error of around 2 mm, which is similar to the results achieved in the present prototypes. Regarding the validation of the prototypes, other medical imaging techniques could be used aside from CT scanning [36]: positron emission tomography (PET), single-photon emission computed tomography (SPECT), magnetic resonance imaging (MRI), ultrasound (US), and mammography. Amongst the previously mentioned medical imaging techniques, the least used for phantom imaging is mammography [36]. CT is shown to be the best due to its good spatial resolution, high contrast and signal-to-noise ratio, which enhance the differentiation of the anatomical structures [37]. Additionally, CT with MRI are the most used for quantitative results measurements; however, Mitsouras et al. [38] concluded that MRI demonstrated larger differences in the phantom compared to the CT data.
The surgical planning prototypes, which were manufactured by printing a mould in PLA using FFF or PA12 using SLS, silicone casting and manufacturing PA12 with SLS for the internal parts, had a total cost of approximately €500. For Madurska et al. [39], manufactured using TangoPlus ® and TangoBlack ® (Stratasys, Rehovot, Israel) without casting or moulding, the cost ranged between USD$500-600; it was approximately USD$1000 in Igami et al. [40] or more than USD$2000 per model in Prashanth et al. [41]. This was more expensive than our prototypes if only the material costs were taken into account. On the other hand, Witowski et al. [24] manufactured using silicone and FFF. In this case, the total cost was around USD$130, which represents a similar cost to our prototype in terms of material costs. The production cost of the mould with SLS technology and PLA material of case #2, as an alternative to the PLA moulds produced by FFF in the other two cases, has been more expensive without providing a significant advantage in the manufacturing process or in the quality of the final result.

Clinical Relevance
The definition and implementation of a digital workflow for pre-surgical simulation training and planning using 3D printing technologies in the treatment of patients with liver cancer represent a promising and unique opportunity. These new techniques allow for a better understanding of complex anatomy and a first approach to 3DP replicas that allow surgical dissection and surgical rehearsal. This opens the way for surgeons in training to better prepare for very complex surgeries. Moreover, although the integration of this new Gels 2023, 9, 339 8 of 15 multidisciplinary approach requires an increased turnaround time in planning, and there is far less experience with soft tissue 3DP than with bone structures, there is some data demonstrating a reduction in surgical time, complications and outcomes [2].

Limitations
The major technical drawback of the manufactured 3D models is that both the blood vessels and the tumour have a rigid consistency. Nevertheless, they were an excellent option for surgeons for preoperative surgical planning and simulation training. This is a new technological development, and research is ongoing in our laboratories to improve the prototypes by using more elastic parts for all anatomical structures. The next step is to combine the soft-tissue-mimicking results with current prototypes [7].

Conclusions
We demonstrate that it is possible to manufacture accurate 3D-printed soft surgical planning prototypes at a low cost compared to the existing alternatives. The 3D models allowed for proper pre-surgical planning and simulation training in the three cases reported, being a valued aid for surgeons. Furthermore, we present a detailed workflow for extending the hospital production of 3D pre-surgical liver models. This full process could also be used in other medical indications or areas, such as neuroblastomas, traumatology or brain tumours. In the future, new materials with advanced properties, such as hydrogels or the combination of different materials (silicones and hydrogels), could also be used for the different anatomical parts, improving tissue mimicking. Another possible approach is the manufacture of multi-material 3D prototypes using a hybrid multi-material 3D printer, in which filaments and slurry-based materials, such as hydrogels and silicones, could be combined.

Cases Presentation
Three paediatric cases were transferred to a major paediatric referral centre for treatment evaluation. All three patients presented complex hepatic tumours. Case #1 was a 2-year-old male. After radiological evaluation using contrast CT (computed tomography) scan and MRI (magnetic resonance imaging) (see the section below), a hepatic mass with biliary tract dilatation was observed. Drainage and a tru-cut biopsy were performed, with a diagnosis of biliary tract rhabdomyosarcoma. Chemotherapy was given according to the recommended SIOP protocol, and after a good response, surgery was advised. Case #2 was a 1-year-old female with a left hepatic mass and alpha-feto protein elevation. Biopsy confirmed Pretex II hepatoblastoma, and surgery was advised. Case #3 was a 1-year-old male presenting a hepatic mass compatible with hepatic hamartoma. (See Table 4). 3D reconstructions and 3D printed models were performed for surgical planning purposes. Table 4 summarises the patients' information.

Digital Workflow for 3D Printing Pre-Surgical Simulation-Based Training
The pre-surgical 3D planning and simulation based-training requires a multidisciplinary team and a cross-functional alignment of surgeon, oncologist, radiologist and 3D planning bioengineer. The summarised process workflow is defined and presented in the flowchart below (See Figure 4):

Digital Workflow for 3D Printing Pre-Surgical Simulation-Based Training
The pre-surgical 3D planning and simulation based-training requires a multidisciplinary team and a cross-functional alignment of surgeon, oncologist, radiologist and 3D planning bioengineer. The summarised process workflow is defined and presented in the flowchart below (See Figure 4): The following sections describe the main stages of the process in more detail.

Image Acquisition
The first step is to obtain information about the anatomy, geometry and tissue composition of the corresponding normal and pathological structures. Radiological multiplanar imaging, such as CT or MRI, provides the most useful information, not just about the findings (e.g., localisation, number of lesions, etc.) but also allows the segmentation of different structures.
The patient's livers were scanned using computed tomography (iCT 256 Philips) with a standard paediatric abdominal CT protocol. In all cases, the abdominal CT was performed following these parameters: 1mm slice thickness with 0.5 reconstruction, 80 kV, exposure modulation and IMR reconstruction (See Figure 5). The CT protocol included iodinated contrast injection (split of standard weight/dose) to obtain in one acquisition a double vascular image (arterial and venous-portal). To complement the CT acquisition, an MRI study was added to evaluate the tissue features. The protocol included T2, DWI and axial post-contrast dynamic 3D T1. The following sections describe the main stages of the process in more detail.

Image Acquisition
The first step is to obtain information about the anatomy, geometry and tissue composition of the corresponding normal and pathological structures. Radiological multiplanar imaging, such as CT or MRI, provides the most useful information, not just about the findings (e.g., localisation, number of lesions, etc.) but also allows the segmentation of different structures.
The patient's livers were scanned using computed tomography (iCT 256 Philips) with a standard paediatric abdominal CT protocol. In all cases, the abdominal CT was performed following these parameters: 1mm slice thickness with 0.5 reconstruction, 80 kV, exposure modulation and IMR reconstruction (See Figure 5). The CT protocol included iodinated contrast injection (split of standard weight/dose) to obtain in one acquisition a double vascular image (arterial and venous-portal). To complement the CT acquisition, an MRI study was added to evaluate the tissue features. The protocol included T2, DWI and axial post-contrast dynamic 3D T1.

Image Segmentation and Surface Reconstruction
The images acquired using the CT and MRI techniques were saved in DICOM (digital imaging and communications in medicine) format ( Figure 5). The image segmentation was carried out by an expert radiologist to extract the anatomy to be used in the 3D planning and printing. A semi-automatic segmentation was performed using the IntelliSpace Portal© software from Philips© (Amsterdam, The Netherlands). The segmentation process helps obtain the 3D surface STL model with the addition of the DICOM segmented mask regions of interest and is exported to an STL file. els 2023, 9, x FOR PEER REVIEW

Image Segmentation and Surface Reconstruction
The images acquired using the CT and MRI techniques were saved in imaging and communications in medicine) format ( Figure 5). The ima was carried out by an expert radiologist to extract the anatomy to be use ning and printing. A semi-automatic segmentation was performed usin Portal© software from Philips© (Amsterdam, Netherlands). The segm helps obtain the 3D surface STL model with the addition of the DICOM regions of interest and is exported to an STL file.
The workflow of the surgical planning prototypes can be seen in Fig The workflow of the surgical planning prototypes can be seen in Figure 6. The different parts of the prototypes are highlighted in different colours so that the different anatomical structures are clearly distinguished. Regarding case #1, the part of the image segmentation of Figure 6A depicts the different anatomical parts: (1) the aorta and hepatic artery are in red, (2) the portal vein is in purple, (3) the vena cava and supra-hepatic veins are in blue, (4) the biliary tract can be found in brown, dilated by tumour obstruction, and finally, (5) green corresponds to the tumour. Case #2 in Figure 6B shows: (1) the portal vein in purple, (2) the vena cava and supra-hepatic veins in blue and (3) the tumour in light blue. Case #3 in Figure 6C shows: (1) the portal vein is in purple, (2) the vena cava and supra hepatic veins are in blue and (3) the tumour is grey.

CAD Design and 3D Moulding
The extracted 3D STL files were transferred to the 3D bioengineering spec

CAD Design and 3D Moulding
The extracted 3D STL files were transferred to the 3D bioengineering specialists in order to create computer-aided design (CAD) files. The 3D models of each case were created using Autodesk Meshmixer©/MeshLab© software (San Rafael, CA, USA). After clinical validation of the 3D model reconstruction by a senior paediatric oncology surgeon and a senior radiologist, the virtual simulation of the procedure and calculation of the potential tumour volumes were made. Next, the preparation of the different anatomical parts to be printed was begun. In these cases, in making the models, it was decided to opt for combining 3D printing and material casting for the liver tissue with rubber-based materials to reduce the cost of the model and better mimic liver parenchyma characteristics.

3D Printing and Silicone Casting of the Phantom
For the 3D printing of the surgical planning prototype, the moulding technique was used, in which a material is cast inside a 3D-printed mould. The embedded inner anatomical parts (vessels, tumours, biliary tract, etc.) were manufactured using 3D printing SLS technology and PA 12 material ( Figure 6). The 3D printer used was a Ricoh AM S5500P at CIM UPC facilities, which has a layer thickness of 0.08-0.1 mm, displayed in high resolution. PA 12 was used since it is the best option in order to place rigid parts, like the inner embedded structures (veins, tumour, biliary tract), precisely inside the mould when the material is cast. Once these inner parts are 3D printed, they are coloured so that the different anatomical structures can be distinguished. Regarding the mould, two different approaches were carried out. For cases #1 and #3, the outer mould was manufactured using a polylactic acid (PLA) filament in FFF. The 3D printer used was a Sigma model (BCN3D Technologies, Barcelona, Spain), which offers a dimensional precision of ± 0.2 mm and can achieve a layer thickness of 25 µm. For case #2, the outer mould was manufactured using PA12 with the SLS technology. The materials, as well as the manufacturing process, were changed in order to show the possibility of using different technologies and assess their impact both economically and in terms of quality of the final model. Once all parts are 3D printed, they are assembled, and the silicone is cast. The commercial silicone used is Essil 291 Resin-38 Shore A at a volume ratio of 10:1 with a catalyst (ESSIL 292 Catalyser) for all cases (see Figure 7). According to Curtis et al. [42], silicone gels are normally supplied in a two-part fluid system and cured through a platinum-catalysed addition reaction. Parts A (in this case Essil 291 Resin) and B (in this case ESSIL 292 Catalyser) are mixed at a desired ratio (in this case 10:1) and cured (usually by exposure to elevated temperature) to yield a sticky but cohesive mass.

Validation
CloudCompare© V2.11 was used for the validation of the surgical planning prototypes. The printed anatomical models were compared against the computer-aided designed models to assess printing accuracy [20]. The printed models were scanned and segmented using the same acquisition and segmentation technique as the original cases, obtaining the STL files. Both STL files of each case (the one obtained from the initial patient acquisition and the one from the 3D-printed model) were aligned by selecting different referential points in each mesh. Then, the distance between both meshes was computed using the cloud-tocloud distance (Hausdorff distance algorithm), which is a dimensional measurement for comparing image segmentations between two set points [43].

Validation
CloudCompare© V2.11 was used for the validation of the surgical planning prototypes. The printed anatomical models were compared against the computer-aided designed models to assess printing accuracy [20]. The printed models were scanned and segmented using the same acquisition and segmentation technique as the original cases, obtaining the STL files. Both STL files of each case (the one obtained from the initial patient acquisition and the one from the 3D-printed model) were aligned by selecting different referential points in each mesh. Then, the distance between both meshes was computed using the cloud-to-cloud distance (Hausdorff distance algorithm), which is a dimensional measurement for comparing image segmentations between two set points [43].  Funding: The research undertaken in this paper has been partially funded by the QuirofAM project (Exp. COMRDI16-1-0011) co-financed by the European Union through the European Regional Development Fund FEDER with the support of ACCIÓ-Generalitat de Catalunya 2014-2020.