In this study, we investigated the application of the three modalities CT, DR, and MRI for age assessment of the knee in the same individual. A comparison of the three modalities reveals that there is good agreement between PMDR and PMCT as well as PMDR and PMMRI, but only moderate agreement between PMCT and PMMRI. Few have done comparative studies dealing with forensic age estimation [13, 29,30,31,32, 37]. Therefore, a comparison of our results to other studies is limited. Nevertheless, this study can shed light on the ossification stages across different modalities and provide additional insight into post-mortem imaging.

The fact that there is merely moderate agreement between PMCT and PMMRI images is not in line with a previous study conducted in our department which concluded that CT and MRI of the clavicles used for forensic age estimation purposes can be used interchangeably [32]. However, this study was on a different bone; it used fewer stages and included only one growth plate, which may result in a higher level of agreement. On the other hand, this partially verifies what other authors have previously claimed, namely, that modality-specific reference data is necessary in the practice of age estimation [29,30,31]. This is further confirmed by the fact that complete agreement for all three modalities was only seen in 13/33, 14/33, and 9/29 cases for the femoral, tibial, and fibular bone, respectively. Overall we observed a higher stage for PMDR compared to PMCT and PMMRI, respectively, which is similar to other studies [20, 29, 30, 37, 38]. Moreover, similar to other studies, we found that epiphyseal fusion/closure occurs earlier on DR images than MRI images [20, 37] and specifically also for the distal femoral epiphysis in males [37]. This may be due to the better contrast and definition provided by MRI [20]. In our study, stage 5 was not observed on PMMRI images but only on PMCT and PMDR images. This is in line with other studies that have applied the staging method by Schmeling et al. and Kellinghaus et al. and with a similar MRI protocol as ours [5, 6, 19, 39]. In those studies, they conclude that the reason must be that stage 5 (disappearance of the epiphyseal scar) lies above their included age limit or the possibility that the epiphyseal scar persists into old age. In our study, stage 5 was observed on several PMCT and PMDR images, whereas it was not observed on PMMRI. This observation indicates that the above-mentioned qualities of the MRI might explain the missing stage 5 in our study and those earlier studies.

The lower intraobserver and interobserver agreement seen for PMCT compared to the other two modalities may be ascribed to the slice thickness. In four cases found in our archive, both 1.0 mm and 2.0 mm PMCT images were available and a comparison of those showed that in more than 50% of the cases, a lower stage was given on 1.0 mm CT images compared to 2.0 mm. This indicates that slice thickness is just as important for the staging of the knee on CT images as it is for the clavicles. The slice thickness was 2 mm in our study, and 1 mm is recommended for the staging of clavicles on CT images [40]. We also noted that our interobserver agreement was not as good as our intraobserver agreement which might be due to the different experience and the qualification of the observers. It has been shown that experience and specific qualification affects the epiphyseal staging of clavicles [41]. In addition, interobserver agreement was generally higher for the distal femoral epiphysis compared to the proximal tibial epiphysis. This has previously been noticed and explained by the thicker shape of the femur compared to the tibia [20].

When comparing our results with other studies using the same staging method and MRI sequence, we found a minimum age of 16 years for stage 3c for the distal femoral epiphysis in males which is similar to the minimum age of 16.13 years obtained by Ottow et al. [19] and 15.8 years obtained by Ekizoglu et al. [39]. For females, the minimum age for stage 4 for the distal femoral epiphysis and proximal tibial epiphysis was 16 years which is similar to the minimum ages 16.2 and 15.6 and 16.13 and 15.87 years obtained by Krämer et al. [5, 6] and Ottow et al. [19], respectively. In our study, the minimum age obtained for stage 4 for the proximal tibial epiphysis was 16 years for males, which compares to the minimum ages 16.3, 15.90, and 15.8 years from the studies of Krämer et al. [5], Fan et al. [37]. and Ekizoglu et al. [39], respectively. However, our minimum age of 21 years for males for stage 4 for the distal femoral epiphysis is much higher than the results from the other studies [6, 19, 37, 39]. This result suggests that stage 4 of the distal femoral epiphysis may be helpful in the estimation of the 18-year age limit as previously proposed by some authors [6, 21] but then disproved by others [19, 37, 39]. However, the validity of the minimum and maximum ages is limited by the low number of cases and the lack of even distribution in terms of age and sex. Indeed, due to the unbalanced age distribution with most of the subjects included above 18 years of age, selection bias needs to be considered. The question of the 18-year-old limit must be verified on a larger population with a more evenly age and sex distribution.

There are limitations to this study which need to be mentioned and considered. As mentioned in the previous paragraph, the low number of subjects included and the unbalanced distribution in both age and sex are evident limitations. Another possible limitation and probably the most important one in this context is the question of whether research on post-mortem individuals is transferable to living individuals. It is well-known that radiological images conducted post-mortem look differently compared to images of living individuals. For instance, MRI images are influenced by body temperature, and CT images by putrefaction [42, 43]. Nevertheless, a previous study comparing clavicle fusion in living and post-mortem individuals showed no statistically significant difference in stages, thereby concluding that post-mortem MRI (PMMRI) and MRI of the living are likely transferable [16]. A final limitation which needs to be mentioned is the fact that the scan protocol of the PMCT images was meant for our routine investigation and not the assessment of knee epiphysis. For instance, we observed a lower in-plane resolution when assessing the images in coronal and sagittal plane using MPR which could have been improved by reconstructing from isotropic voxels. Furthermore, we could have achieved a higher in-plane resolution on axial images by adjusting the FOV for our axial scan.

On the other hand, a strength of our study is that we have assessed the images in more than one projection or plane including all slices and all three bones of the knee, while many other studies involving epiphyseal growth staging of the knee have chosen a specific region of the bone to assess and only in one plane or projection [5, 6, 13, 19, 22, 23, 37, 39, 44, 45]. Though, some authors advise against the assessment of the proximal fibular epiphysis on MRI due to its lack of visibility in all slices and therefore might see its inclusion in this study as a limitation rather than a strength [19]. This methodological approach was applied in order to adhere to the benefit of the doubt principle. Assessing all planes/projections and all slices guarantees that uncertainties in the staging process can be accounted for. Nevertheless, assessment of all slices in more than one plane or projection could perhaps also explain the disagreement between observers and modalities since the more images there are to assess the more risk of disagreement/doubt on the stage to assign.

Variations both across different bones and among different individuals make it difficult to find a method for forensic age estimation that is suitable for all individuals and in any case. Therefore, studies on age estimation differ in their choice of study setup including the staging method. The combined classification by Schmeling et al. and Kellinghaus et al. that was applied in this study was developed for the clavicles using CR and CT, respectively. Various studies have shown the reproducibility and feasibility of this staging method on CR, CT, and MRI images [5, 29, 41]. We, therefore, chose to apply this method in our study and believed it was the best choice for comparison of modalities than any of the other staging methods available.