A multi-method assessment of 3D printed micromorphological osteological features

The use of 3D models and prints in forensic science can be beneficial for demonstration purposes in a courtroom setting [1, 2]. 3D printed osteological replicas have been examined for their accuracy [3, 4] and are increasingly being used in a wide range of forensic applications [5‐10]. Osteological samples can present with fine micromorphological surface details that can be important for evaluative interpretation of forensic events and external processes such as from trauma, taphonomy, and pathology. The presentation of micromorphological details can therefore enhance understanding of, for example, mechanisms of injury; however, previous research using CT data has demonstrated difficulty in obtaining accurately 3D printed micromorphological details [11]. This study therefore investigated the accuracy of 3D printed bones that exhibited a range of micromorphological surface details (approximately less than 1 mm), utilising micro-computed tomography (micro-CT) scanning using a multi-method experimental approach that included a computational mesh comparison.

A preliminary investigation into the accuracy of 3D printing bone replicas from CT scans by Carew et al. [4], tested six different 3D printers and identified a range (± 2.0 mm) that printed reconstructions should fall within to be considered accurate, this range originated from the generally accepted range for error in forensic anthropology [12]. Additionally, a number of recommendations for obtaining accurate prints were offered, including the use of Selective Laser Sintering (SLS) [4]. Subsequent research has also assessed the surface quality of printed bones using SLS printing and found that not all surface details were well represented [11]. The complex nature of bone structure and highly textural features of bone surfaces can be challenging to retain at sufficient levels in 3D printed replicas. Carew et al. [11] suggested that micro-CT may be a more suitable digital capture method for scanning bones where high surface detail is required, due to the greater spatial resolution obtained from micro-CT scanning and its ability to capture fine details, such as cut marks [3].

Micro-CT is a non-invasive, transmissive scan method that captures volumetric data using x-radiation of a sample that rotates within the scan chamber [13]. Its value in casework has been presented for histological analysis such as species identification [14], for soft and hard tissue examination in forensic pathology [15], for physical fit analysis of bone fragments [16], and for cutmark analysis [3, 17]. The analysis of micromorphological details is important for evaluative interpretation in forensic casework in terms of indicating possible trauma (e.g., cutmarks or striations), taphonomy (e.g., heat induced bones fractures), or pathology (e.g., evidence of disease or malnutrition). Micro-CT has the potential to capture these details due to its high spatial resolution and greater image quality compared to clinical CT [18].

The detailed capture and representation of fine details from trauma pose a new challenge for crime reconstructions. Cranial fracture lines have been replicated through 3D printing with limited success; studies have found that fractures from blunt force trauma captured using surface scanning techniques [19] and from gunshot trauma from post-mortem computed tomography [20] could only be partially observed. Experimental research using micro-CT to capture cutmarks on porcine bone found the 3D printed replica bones to be accurate to the original bones to < 0.62 mm for overall morphology and < 0.36 mm for toolmarks [3]. The models were 3D printed using three printers (Polyjet Material Jetting printer, Fused Deposition Modelling (FDM), and Stereolithography (SLA)), and rescanned using a surface scanner to perform a mesh comparison between the original models and printed models. Baier et al. [3] demonstrate the utility of the mesh comparison method for providing good quantitative results, but also offer caution against 3D printing important features that are smaller than 3 mm.

Little research has been identified that has examined 3D printed bones with taphonomy; however, 3D printed burnt bone fragments provided a useful tool for demonstrating the reconstruction and physical fit of fragments, which was also non-destructive and valuable for use with fragile fragments [16]. Both ante-mortem healed trauma and pathological lesions can also be important intelligence for identification purposes; however, few studies have investigated the printing of pathological specimens for crime reconstruction purposes. Intricate surface details from a calcified leiomyoma (a smooth muscle tumour) were reported to be well presented on a 3D printed replica [21]. The leiomyoma was captured using surface scanning and reconstructed using FDM printing, with the pathological changes appearing to be a good representation on the 3D print from the raised areas, but the recessed areas were not well captured due to the use of surface scanning [21]. Similarly, a scaled-up printed replica femur from a 19-week-old foetus that presented with osteogenesis imperfecta was reported to be a useful tool for demonstrating pathological features [22]. However, the use of scaling has not been investigated in 3D printing for crime reconstruction purposes.

While there is currently limited published research addressing the use of 3D printed bones exhibiting indicators of trauma, taphonomy, and pathology, the value of such prints has been highlighted [16, 19, 21, 22]. The accuracy of 3D printed bones is key to ensure that accurate and representative prints are produced, particularly if they were to be used within a court of law [3, 4, 11]. Printing errors can arise through the scanning, modelling, and/or printing steps and therefore each step used ought to be carefully considered and justified prior to implementation [20, 23]. Studies have investigated the effect of individual parameters in bone modelling and printing, such as during scanning [24], thresholding and segmentation [25], and generally for comparing medical models [23]. To ensure that printed replicas are of sufficient quality for the level of interpretation required by the expert or the jury, steps can be implemented to both mitigate against print errors and quantify print accuracy [3, 11]. Mesh comparison methods offer a comparison workflow that is independent of the observer measurement error found with manual measurements and provides a higher quantitative method for assessing print accuracy [3, 26]. Therefore, this study sought to assess the surface quality of 3D prints exhibiting alterations from trauma, taphonomy and pathology processes, to determine whether it was possible to observe accurate micromorphological details on 3D printed bones.

A multi-method experimental design was developed to evaluate the surface quality and accuracy of 3D printed bones that exhibited intricate surface details, utilising four approaches that allowed for a robust examination of the overall print accuracy. The multi-method assessment included the following four steps:

A single experienced observer performed each of the four assessment steps. The observer was a UK Level III Certified Forensic Anthropologist with 5 years’ experience of generating and assessing 3D printed bones. A single observer was used to reduce the possibility of observer bias and different experience levels from multiple observers affecting the results.

Each of the nine bones was imaged using micro-CT, a 3D model generated from the scan data, and then an SLS 3D print generated from each model (Fig. 2).

Bild vergrößern

Verification of the accuracy of 3D printed replicas is an important step in crime reconstructions. Moreover, the representation of intricate details from process such as trauma, taphonomy and pathology needs to be reliable for evaluative interpretation and courtroom demonstration purposes. This multi-method experimental approach study utilised nine bone samples exhibiting various types of micromorphological details. A preliminary scan parameter validation test was conducted to validate the use of the Optimise setting for the micro-CT scans which showed very low distance values between the models scanned at varying parameters through a mesh comparison. Adjusting the scan parameters did not have an observable effect on the size of the 3D models, which therefore justified the use of the ‘Optimise’ setting for the subsequent bone imaging.

A quality control step (step 1) examined the overall size of the 3D printed bones. The 3D prints were identified to be within ± 1.0 mm of the original bone 3D models, with no statistically significant result found between the manual and virtual measurement data. These results established the overall accuracy of the 3D printed bones and confirmed that no scaling error had occurred. These results were congruent with those obtained by Carew et al. [11] who used a similar quality control technique, although using a clinical CT scanner not micro CT. The qualitative assessment (step 2) established that fracture lines were observable on the 3D printed surfaces, but that these lines often presented as continuous grooves, rather than discontinuous separate margins. Exposed trabecular bone and toothmarks were also generally well-presented on the prints, but with loss of detail or depth often identified. Generally, the visual comparison found that the 3D printed bone replicas adequately represented the micromorphological details, whereby the features were usually observable. However, there was often loss of detail, depth, and fine texture, such as the shine from eburnation found on bone G (the patella). Factors such as the scan parameters, the level of smoothing applied during modelling and the resolution of the printer will all influence the presentation of the micromorphological details. As discussed, each of these needs careful consideration and further research to explore and quantify these effects.

Additionally, it was noted that the contrast between the colour, texture and depth was sometimes missing on the 3D prints, primarily due to the replication using a monochrome white material. Shading and colour cues are necessary for monocular vision processing [32], but this factor cannot be replicated with monochrome prints. It may be possible to produce multicoloured 3D prints using photorealistic printing from photographs coupled with surface or volumetric scanning [31], however, this would only be an accurate colour representation if captured from dry or exposed human remains. Further, this photorealistic aspect could mean that the 3D printed replicas become too realistic or too disturbing, as with colour photographs [33, 34]. It is therefore proposed that perhaps the loss of detail found with a monochrome 3D printed replica is an accepted loss that facilitates the use of the 3D print as a less-graphic, sanitised option for courtroom demonstrations.

An objective visual comparison (step 3) utilised the surface scoring method developed by Edwards and Rogers [19] to quantitatively score the surface quality of the printed bones. All of the 3D printed bones obtained the highest score of 5 for the general morphology category, which corresponds to the quality control step, that each of the prints had accurate overall morphology. High scores were also assigned to the detailed morphology category, but with lower scores assigned to the texture and fracture pattern categories. These findings echo those presented by Carew et al. [11], who also found 3D prints to be accurate overall, but limited by loss of finer details. This visual comparison identified that five of the nine bones (D to H) exhibited accurate surface representations overall. Bone B was scored as accurate when the altered method excluding fracture pattern assessment was used. Previous research by Carew et al. [11] utilising clinical CT scanning, identified only two of their nine prints to be accurate from this scoring method. Therefore, this study presents potential improvement in accuracy rates, with five of the nine prints scored as accurate using the micro-CT scanning technique; this is despite the bones used in this study having more complex micromorphology than the previous study. The scoring method worked well as an adaptable objective recording method. However, the terminology ‘resemblance’ in the detailed morphology category is somewhat ambiguous, and left room for dissimilarity.

Finally, the quantitative assessment (step 4) used a mesh comparison to assess the accuracy of the digitised 3D print models. The initial alignment process produced low RMS values, indicating close alignments between the tested models. The scanned 3D printed bones were then found to be accurate to the original bone models to within the ideal threshold of ± 1.0 mm. Moreover, the prints were accurate to within 0.2 mm using the mean distance values indicating a high degree of accuracy. This result is similar to other mesh comparison results reported between 3D bone models, such as Smith et al. [35] (0.1 ± 0.1), Baier et al. [3] (less than 0.62 mm), and Robles et al. [29] (± 0.4 mm). Conversion of the mesh distance values to percentages provided non-dimensionalised values to indicate the accuracy of the prints. The percentage mean differences were approximately 0–1% for length and 0–5% for width. These values again indicated low differences and good congruency between the original bone models and the digitised 3D prints, similar to those previously reported [3, 12]. These values are independent of the size of the sample and provide a valuable figure for forensic reporting purposes that may be more meaningful for lay interpretation.

The majority of the mesh distance values reported were also identified as coming from the internal bone structure, not from the bone surface. Furthermore, while the mesh comparison method is a simple, effective method for comparing model dimensions, the method does not effectively identify differences in surface quality. As seen in this study whereby high model accuracy was obtained, but differences were also seen in the presentation of micromorphological surface details. These findings concur with those from Edwards and Rogers [19], who note that distance values do not necessarily reflect the quality of printed replica, as any morphological differences will be minimal irrespective of digitisation method used. The implementation of visual inspection methods is vital for proper examination of print quality, and mesh comparison methods alone should not be used.

The multi-method experimental approach employed in this study has generated insights that contribute towards the articulation of a pathway for assessing the accuracy of the surface of 3D printed bones. The results identified that:

The pathway to check for print quality can be used in future crime reconstruction approaches to protect the integrity of the final 3D printed replicas for forensic applications. It should be considered in the wider forensic science framework, considering factors such as accreditation, certification, standardisation, and the guidance from the Forensic Science Regulator in the UK. This presents an opportunity to develop systems for producing robust printed reconstructions for forensic applications. These findings complement the growing body of published literature supporting the use of 3D printed bones in forensic anthropology crime reconstructions [3‐5, 11, 16, 20] and can be incorporated into emerging forensic 3D printing guidance.