3D printing as a pedagogical tool for teaching normal human anatomy: a systematic review

Background Three-dimensional-printed anatomical models (3DPAMs) appear to be a relevant tool due to their educational value and their feasibility. The objectives of this review were to describe and analyse the methods utilised for creating 3DPAMs used in teaching human anatomy and for evaluating its pedagogical contribution. Methods An electronic search was conducted on PubMed using the following terms: education, school, learning, teaching, learn, teach, educational, three-dimensional, 3D, 3-dimensional, printing, printed, print, anatomy, anatomical, anatomically, and anatomic. Data retrieved included study characteristics, model design, morphological evaluation, educational performance, advantages, and disadvantages. Results Of the 68 articles selected, the cephalic region was the most studied (33 articles); 51 articles mentioned bone printing. In 47 articles, the 3DPAM was designed from CT scans. Five printing processes were listed. Plastic and its derivatives were used in 48 studies. The cost per design ranged from 1.25 USD to 2800 USD. Thirty-seven studies compared 3DPAM to a reference model. Thirty-three articles investigated educational performance. The main advantages were visual and haptic qualities, effectiveness for teaching, reproducibility, customizability and manipulability, time savings, integration of functional anatomy, better mental rotation ability, knowledge retention, and educator/student satisfaction. The main disadvantages were related to the design: consistency, lack of detail or transparency, overly bright colours, long printing time, and high cost. Conclusion This systematic review demonstrates that 3DPAMs are feasible at a low cost and effective for teaching anatomy. More realistic models require access to more expensive 3D printing technologies and substantially longer design time, which would greatly increase the overall cost. Choosing an appropriate image acquisition modality is key. From a pedagogical viewpoint, 3DPAMs are effective tools for teaching anatomy, positively impacting the learning outcomes and satisfaction level. The pedagogical effectiveness of 3DPAMs seems to be best when they reproduce complex anatomical areas, and they are used by students early in their medical studies.


Introduction
Practiced since Ancient Greece on animals, cadaver dissection is one of the main methods used to teach anatomy.Cadaveric dissection, carried out during hands-on training, supports the theoretical lessons given to medical students in universities and is currently considered the gold standard for learning anatomy [1][2][3][4][5].However, there are many obstacles to using human cadaveric specimens, prompting a search for new pedagogical tools [6,7].Some of these new tools are extended reality, digital tools, and 3D printing.According to a recent literature review by Santos et al. [8] on the value of these new technologies for teaching anatomy, 3D printing appears to be one of the most relevant resources both in terms of its educational value to students and the feasibility of its implementation [4,9,10].
3D printing is not new.The first patents related to this technology date back to 1984: A Le Méhauté, O De Witte and JC André in France and 3 weeks later, C Hull in the USA.Since then, this technology has undergone continuous development, and its use has spread to numerous fields.For example, NASA printed the first object outside the planet Earth in 2014 [11].The medical field has also appropriated this new tool, thus reinforcing the desire to develop personalized medicine [12].
Many authors have demonstrated the pedagogical benefits of using 3D-printed anatomical models (3DPAM) for medical education [10,[13][14][15][16][17][18][19].When it comes to teaching human anatomy, non-pathological and anatomically normal models are required.Several reviews have studied pathological models or training models for a medical/surgical procedure [8,20,21].With the intention of developing a hybrid teaching model for human anatomy that incorporates new tools such as 3D printing, we carried out a systematic review to describe and analyse how 3D-printed objects made for teaching of human anatomy are created and how students evaluate the pedagogical contribution of these 3D objects.

Materials and methods
This systematic review of the literature was conducted in June 2022 without time limitation using the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines [22].

Eligibility criteria
Inclusion criteria were all research papers dealing with 3DPAM in anatomy teaching/learning.Literature reviews, letters, or articles studying pathological models, animal models, archaeological models, and medical/surgical training models were excluded.Only articles published in English were selected.Articles without available online abstracts were excluded.Articles dealing with several models -at least one of which was anatomically normal or had trivial pathology that did not alter the pedagogical value -were included.

Search strategy
A literature search was performed in the electronic Pub-Med database (National Library of Medicine, NCBI) to identify relevant studies published up to June 2022.The following search terms were used: education, school, learning, teaching, learn, teach, educational, threedimensional, 3D, 3-dimensional, printing, printed, print, anatomy, anatomical, anatomically, and anatomic.OR anatomic[Title/ Abstract]).Additional articles were identified through a manual search in the PubMed database and by looking through the references of other scientific articles.No date restriction was applied but the "human" filter was used.

Study selection
All retrieved titles and abstracts were screened against the inclusion and exclusion criteria by two authors (EBR & AL), and any study that did not meet all the eligibility criteria were excluded.Full-text publications of the remaining studies were obtained and screened by three authors (EBR, EBE & AL).Any disagreement in the selection of articles was resolved, if necessary, by a fourth person (LT).Publications that met all the inclusion criteria were included in this review.

Data extraction was performed independently by two authors (EBR & AL) and supervised by a third (LT).
The extracted data consisted of: --study characteristics: publication date, country of authors, type of study --model design data: anatomical region, specific anatomical part, initial model used for 3D printing, acquisition method, segmentation and modelling software, type of 3D printer, type and number of materials, printing scale, colours, cost of printing --morphological evaluation of the model: model used for comparison, medical evaluation by an expert/ teacher, number of raters, type of evaluation --pedagogical performance of the 3D model: student knowledge assessment, assessment methods, number of students, number of comparison groups, randomization of students, type of education/student --advantages and disadvantages.
All data were extracted in predefined forms.

Study selection
Four hundred eighteen studies were identified in the MEDLINE database; 139 articles were excluded by the "human" filter.After the title and abstract were analysed, 103 studies were selected for reading of the full text.Thirty-four articles were excluded because they were either pathological models (9 articles), medical/surgical training models (4 articles), animal models (4 articles), 3D radiology models (1 article) or were not original scientific papers (16 articles).A total of 68 articles were included in this review.Figure 1 summarizes the selection process with a flowchart.

Study characteristics
All studies were published between 2014 and 2022, with the average year of publication being 2019.Of the 68 articles included, 33 (49%) studies were descriptive and experimental, 17 (25%) were purely experimental and 18 (26%) were purely descriptive.Among the 50 (73%) experimental studies, 21 (31%) used randomization.Only 34 studies (50%) included a statistical analysis.Table 1 summarizes the characteristics of each study included.
Regarding the original model or file used for designing the 3DPAM, 23 articles (34%) mentioned the use of patient data, 20 articles (29%) the use of cadaver data, 17 articles (25%) the use of a database, and 7 studies (10%) did not disclose the origin of the file used.
In 47 studies (69%), the 3DPAMs were designed from CT scans, while 3 studies (4%) specified using micro-CT scans.In 7 articles (10%), the 3D objects were designed from optical scanners, in 4 articles (6%) from MRI and in 1 article (1%) from a camera and microscope.In 14 articles (21%), the origin of the source files for the design of the 3D model was not mentioned.The average spatial  resolution was less than 0.5 mm for creating the 3D files.
Thirty-four articles (50%) mentioned a cost.Nine articles (13%) mentioned the cost of the 3D printer and the raw materials.Printers ranged in price from 302 USD to 65,000 USD.The cost per model, when specified, ranged from 1.25 USD to 2800 USD; these extremes corresponded to a bone specimen [47] and a high-fidelity retroperitoneal model [48].Table 2 summarizes the model design data for each included study.

Discussion
This systematic review summarizes 68 articles published over 9 years, highlighting the scientific community's interest in 3DPAM as a pedagogical tool for teaching normal human anatomy.Every anatomical region has been studied and printed in 3D.Among these articles, 37 compared the 3DPAM to another model and 33 evaluated the pedagogical relevance of the 3DPAM for students.
Given the differences in the design of studies on 3D printing in anatomy, we did not feel it was appropriate to carry out a meta-analysis.A meta-analysis published in 2020 focused mainly on post-training tests of anatomical knowledge, without analysing the technical and technological aspects of the design and manufacture of 3DPAMs [10].

Model design data
The cephalic region was the most studied, probably because its anatomical complexity makes it difficult for students to picture this anatomical region in 3D space, compared to the limbs or trunk.CT scan was by far the most used image acquisition modality.This modality is widely available, especially in healthcare facilities, but its spatial resolution is limited, and its soft-tissue contrast is low.These limitations make CT scan unsuitable for segmentation and modelling of the nervous system for example.On the other hand, CT scan was preferred for   the segmentation/modelling of bone tissue; the bone/soft tissue contrast facilitates these steps before 3D printing of an anatomical model.Micro-CT, on the other hand, was cited as the reference technology in terms of spatial resolution for the acquisition of bone tissue images [70].
An optical scanner or MRI can also be used for image acquisition.Higher resolution prevents the smoothing of bone surfaces and preserves the subtleties of the anatomy [59].The choice of models also influences the spatial resolution; for example, plastinated models have lower resolution [45].A graphic designer was needed when creating highly customized 3D models, which increases the cost (25 to 150 USD per hour of work) [43].Obtaining a good quality.STL file was not sufficient to produce a good quality anatomical model.The printing parameters such as the orientation of the anatomical model on the printing plate must be defined [29].Some authors suggested that advanced printing technologies such as SLS should be used whenever possible to improve the 3DPAM's accuracy [38].The help of a professional was required to make the 3DPAM; the most requested professionals were an engineer [72], radiologist, [75] graphic designer, [43] and anatomist [25,28,51,57,76,77].Segmentation and modelling software are important factors for obtaining an accurate anatomical model, but the price of these software packages and their complexity hinder their use.Some studies compared the use of different software packages and printing technologies, highlighting the advantages and disadvantages of each [68].In addition to modelling software, printing software is required that is compatible with the chosen printer; some authors preferred to use online 3D printing [75].If enough 3D objects will be printed, the investment may be financially profitable [72].
Plastic was by far the most used material.It is the material of choice for 3DPAM due to its large range of textures and colours.Several authors praised its high strength compared to traditional cadaveric or plastinated models [24,56,73].Some plastics even have flexural or tensile properties.For example, the Filaflex used with FDM technology can stretch up to 700%.For some authors, it is the material of choice for reproducing muscles, tendons and ligaments [63].On the other hand, two studies raised questions about the direction of the fibres as printed.Indeed, the direction of the muscle fibres is critical when modelling a muscle, along with its insertions, innervation and function [33].
Surprisingly, few studies mentioned the printing scale.Since many consider a 1:1 scale as standard, the authors may have decided not to mention it.The possibility of enlargement has not been explored much despite its benefit for directed teaching in large groups, especially given the increasing number of students per class where the actual size of the model is an important element.Of course, a full-size scale makes it easier to locate the various anatomical elements and to transpose it to patients, which probably explains why this scale is often used.
Among the multiple printers available on the market, those that provide high-definition printing in colour and in several materials -thus several textures -using Pol-yJet technology (material jetting or binder jetting) cost between 20,000 and 250,000 + dollars (https:// www.aniwaa.com/).This high cost likely restricts the diffusion of 3DPAMs in medical schools.In addition to the price of buying a printer, the materials needed for material jetting cost more than those used for SLA or FDM printers [68].The price of SLA or FDM printers is also more manageable, ranging from 576 to 4999 € in the articles listed in this review.According to Tripodi and colleagues, bone parts could be printed for 1.25 USD each [47].Eleven studies concluded that 3D printing costs less than plastinated or commercial models [24,27,41,44,45,48,51,60,63,80,81,83].Furthermore, these commercial models are intended for patient information and do not have sufficient detail to be used for teaching anatomy [80].These commercial models were considered inferior to 3DPAMs [44].It is important to note that -in addition to the printing technology used -the final cost is also proportional to the scale and thus the final size of the 3DPAM [48].For these reasons, the preferred scale was full size [37].

Morphological evaluation of 3D models
Only one study compared a 3DPAM to a commercially available anatomical model [72].Cadaveric specimens were the most used comparator for 3DPAM.Despite its drawbacks, the cadaveric model remains a valuable tool for teaching anatomy.A distinction needs to be made between cadaveric dissection, prosections and dry bones.Two studies found that 3DPAMs were significantly more effective than plastinated prosections based on learning tests [16,27].A single study compared one hour of learning using a 3DPAM (lower limb) with one hour of dissection on the same anatomical area [78].There was no significant difference between the two teaching methods.It is likely that few studies have been done on this topic because this comparison is difficult to set up.Dissection by students is a time-consuming task to prepare for.Several dozens of hours of dissection are sometimes necessary, depending on the dissection subjects.A third comparison can be made with dry bones.The studies by Cai and Smith found significantly better test results for the groups who used 3DPAM [51,63].Chen and colleagues specified that students who used the 3D model were better at recognizing structures (skull) but that there was no difference in MCQ results [69].Finally, Tanner and colleagues demonstrated better post-test results for the group using a 3DPAM of the pterygopalatine fossa [46].This literature review identified other new teaching tools.Among the most common were augmented reality, virtual reality, and serious gaming [43].According to Mahrous and colleagues, the anatomical model preference depends on the number of video game hours played by the student [31].On the other hand, the main pitfall of new tools in anatomy education is haptic feedback, especially for virtual-only tools [48].

Pedagogical performance of 3D models
A knowledge pre-test was used in most studies evaluating new 3DPAMs.These pre-tests help to avoid assessment bias.Some authors excluded all students who scored above average on the pre-test before conducting their experimental study [40].Among the assessment biases, Garas and colleagues cited the colouring of the models but also the choice of volunteers among the student classes [61].Staining makes anatomical structures easier to identify.Chen and colleagues imposed strict experimental conditions, with no initial intergroup differences and as much blinding as possible [69].Lim and colleagues suggest avoiding assessment bias by having the post-test assessment prepared by a third person [16].Some of the studies used Likert scales to assess the 3DPAM's appropriateness.This tool is suitable for evaluating satisfaction but nevertheless has important biases that one must be aware of [86].
The educational relevance of 3DPAMs was evaluated mostly in medical students, including first-year students in 14 of the 33 studies identified.In their pilot study, Wilk and colleagues reported that medical students felt 3D printing should be incorporated into their learning of anatomy [87].Eighty-seven percent of students surveyed in the Cercenelli study felt that their second year was the best time to use 3DPAMs [84].Results from Tanner and colleagues also showed that students were better if they had never studied the area [46].These data suggest that the first years of medical school are the best time to incorporate 3DPAMs into the teaching of anatomy.Ye's meta-analysis corroborates this idea [18].Of the 27 articles included in their study, there was a significant difference in test results in favour of 3DPAMs versus conventional models for medical students but not for residents.
3DPAMs were effective as pedagogical tools in terms of achievement, [16,35,39,52,57,63,69,79] long-term knowledge retention [32] and student satisfaction [25,45,46,52,57,63,66,69,84].Expert panels have also been found these models useful [37,42,49,81,82] and two studies highlighted teacher satisfaction with 3DPAMs [25,63].Among all resources, Backhouse and colleagues judged 3D printing to be the best alternative to conventional anatomical models [49].In their first meta-analysis, Ye and colleagues affirm that the post-test results of students who received instruction incorporating 3DPAMs were better than those who received instruction in 2D or on a cadaver [10].However, they did not differentiate the 3DPAMs by their complexity but simply as heart, nervous system and abdomen.In seven studies, 3DPAMs were not superior to other models based on the knowledge tests given to students [32,66,69,77,78,84].In their meta-analysis, Salazar and colleagues conclude that the use of 3DPAMs specifically improves the understanding of complex anatomical structures [17].This concept is consistent with a letter to the editor by Chytas [88].Certain anatomical areas that are considered less complex would not require the use of 3DPAMs, while more complex anatomical areas such as the neck or nervous system would be a reasonable choices for 3DPAMs.This notion probably explains why some 3DPAMs have not been judged superior to conventional models, especially since the model's effectiveness seems to be better when the student has no knowledge in the field.Consequently, a simple model, presented to students who already have some knowledge of the subject (advanced medical students or residents), would be useless for improving student results.

Advantages and disadvantages
Of all the educational benefits listed, 11 studies highlighted the visual or tactile qualities of their models, [27,34,44,45,48,50,55,63,67,72,85] while 3 studies emphasized the strength and durability (33, 50-52, 63,79,85,86).Other advantages were that the students could manipulate the structures, the teacher could save time, they were easier to preserve than a cadaver, the design could be completed in less than 24 h, it could be used as a home study tool and it could be used to teach large groups [30,49,60,61,80,81].The 3D printing of multiple copies for teaching anatomy in large groups, make the 3D printing of models more cost-effective [26].Using 3DPAMs increased mental rotation ability [23], and improved interpretation of cross-sectional imaging [23,32].Two studies found that students exposed to 3DPAMs were more attracted to surgery [40,74].Metal connectors can be incorporated to produce the motion needed to study functional anatomy [51,53] or to print the model with a page-turning design [67].
3D printing made it possible to create adjustable anatomical models by improving certain aspects during the modelling stage, [48,80] creating a suitable base, [59] merging multiple models, [36] using transparency, (49) colour, [45] or making certain internal structures visible [30].Tripodi and colleagues used modelling clay to supplement their 3D printed bone models, highlighting the value of co-creating the model as a teaching tool [47].In 9 studies, colour was applied after printing, [43,46,49,54,58,59,65,69,75] but only once by the students [49].Unfortunately, that study did not assess the pedagogical quality of the model or the teaching sequence.This is something to take into consideration in the context of anatomy education, since the benefits of hybrid learning and co-creation [89] are well known.In response to growing promotions, self-learning has been implemented several times to evaluate models [24,26,27,32,46,69,82].
One study considered the colours of the plastic materials too bright, [45] another that the model was too fragile, [71] and two others pointed out the lack of anatomical variability when a single model was designed [25,45].Seven studies concluded that the anatomical detail was insufficient in their 3DPAM [28,34,45,48,62,63,81].
The segmentation and modelling time was considered very long and the cost very high (about 2000 USD) for more elaborate anatomical models of large and complex regions such as the retroperitoneum or the cervical region [27,48].In their study, Hojo and colleagues specified that it took 40 h to create their pelvic anatomical model [42].The longest segmentation time was 380 h in the study by Weatherall and colleagues where several models were merged to make a finished paediatric airway model [36].Segmentation and printing time was considered a drawback in nine studies [36,42,57,58,74].However, 12 studies criticized the physical properties of their model, particularly its consistency, [28,62] lack of transparency, [30] fragility and unicolor nature, [71] absence of soft tissues [66] or lack of detail [28,34,45,48,62,63,81].These drawbacks could likely have been overcome with more segmentation or modelling time.Loss of acquisition-related information was an issue for three teams [30,74,77].Patient data was used in which the iodinated contrast agent did not provide an optimal view of the blood vessels due to dose limitations [74].The injected cadaveric model appears to be an ideal approach, freeing itself from the "as low as reasonably achievable" principle and limitations in the dose of contrast agent injected.

Limitations
Unfortunately, many articles did not mention certain key features of their 3DPAM.Less than half of the articles specified whether their 3DPAM was coloured or not.The printing scale was not consistently reported (43% of articles) and only 34% of articles mentioned the use of multiple materials.These printing parameters are crucial because they influence the 3DPAM's pedagogical properties.Most of the articles did not provide enough information about the complexity of obtaining the 3DPAM (design time, qualifications of people, cost of software, cost of printing, etc.).This information is essential and must be taken into consideration before thinking about starting a project to develop a new 3DPAM.

Conclusions
This systematic review demonstrates that the design and 3D printing of a normal anatomical model is feasible at a low cost, particularly by using FDM or SLA printers and inexpensive single-color plastic materials.These basic models can nevertheless be improved by adding colour, or adding structures made of various materials.More realistic models -printed with several materials of different colours and textures to reproduce the haptic qualities of the reference cadaveric model as closely as possible -require access to more expensive 3D printing technologies and substantially longer design time.This would greatly increase the overall cost.No matter the chosen printing process, selecting the appropriate imaging modality is key to successful 3DPAMs.The higher the spatial resolution, the more the model will match reality and be usable at advanced levels of study.From a pedagogical point of view, 3DPAMs are effective tools for teaching anatomy, as evidenced by knowledge tests carried out with students and by the students' satisfaction.The pedagogical effectiveness of 3DPAMs seems to be best when they reproduce complex anatomical areas, and they are used by students early in their medical studies.

Fig. 1
Fig. 1 Flow diagram summarizing the identification, screening and inclusion of articles for this systematic review Photopolymers and translucent elastomersresins (Vero Yellow), Tango Plus and Vero-Magenta Abbreviations: ENT Ear nose throat, MCQ Multiple choice question, N No, NS Not specified, Y Yes

Table 1
Summary of study characteristics

Table 1
(continued) Abbreviations: 3DPAM 3D printed anatomical model, N no, Y yesTable 2 Summary of model design data

Table 2
(continued) Abbreviations: AP A posteriori, BJ Binder jetting, ABS Acrylonitrile butadiene styrene, CT Computed tomography, FDM Fused deposition modelling, GBP Great Britain Pound, HIPS High impact polystyrene, micro-CT Microcomputed tomography, MJ Material jetting, MRI Magnetic Resonance Imaging, N No, NS Not specified, PA Polyamide, PLA Polylactic acid, SLA Stereolithography, USD United States dollars, Y Yes

Table 3
Summary of how the morphology of the 3D models was evaluated

Table 3
(continued) AR Augmented reality, CT Computed tomography, NS Not specified, VR Virtual reality

Table 4
Summary of how the pedagogical performance of 3D models was evaluated