Micro‐CT guided illustration of the head anatomy of penguins (Aves: Sphenisciformes: Spheniscidae)

Abstract The illustration is an important tool to aid in the description and understanding of anatomy, and penguins (Aves: Sphenisciformes: Spheniscidae) are an important clade in environmental monitoring, paleontology, and other research fields. Traditionally, anatomic illustration has been informed by dissection. More recently, micro‐computed tomography (micro‐CT) has proven to be a powerful tool for three‐dimensional anatomic imaging, although larger specimens are more challenging to image due to increased X‐ray attenuation. Here, we used traditional dissection and micro‐CT to illustrate the skulls of Aptenodytes patagonicus, Eudyptula minor, and Pygoscelis papua, and the extracranial soft tissue of E. minor. Micro‐CT prevented the loss of orientation, disarticulation, and distortion of bones that might result from cleaning and drying skulls, while immobilization was achieved by freezing the specimens before imaging. All bony elements in the head were accurately depicted. Fixing, dehydrating, and diffusion staining with iodine (diceCT) enabled the identification of muscles and other large nonmineralized structures, but specimen preparation precluded the ability to show smaller nerves and vessels. The results presented here provide a guide for anatomic studies of penguins and our summary of sample preparation and imaging techniques are applicable for studies of other similarly sized biological specimens.

The illustration is an important tool to aid in the understanding of anatomy and numerous illustrations of avian head anatomy are readily available (Baumel et al., 1993;Ghetie et al., 1976;Sosa & Acosta Hospitaleche, 2018). Anatomic studies of the penguin head include descriptions of the skull and the muscles of the jaw (Zusi, 1975), the supraorbital gland of Pygoscelis penguins (Herbert, 1975), the skull and musculature of the emperor penguin (Sosa & Acosta Hospitaleche, 2018) and the skulls of Pygoscelis penguins (Acosta Hospitaleche & Tambussi, 2006). Detailed comparative anatomic descriptions of the head are also available for both extinct and extant penguins Tambussi et al., 2015;Triche, 2007). However, neither illustrated anatomic diagrams of the head comparable to those found in human anatomic textbooks nor micro-computed tomography (micro-CT) renders of the head of the penguin are available, despite their usefulness for didactic purposes or in the veterinary care of these animals.
Micro-CT has emerged as a powerful imaging tool for natural history specimens and comprehensive head anatomy of birds including the dove (Columba livia; Jones et al., 2019) and common buzzard (Buteo buteo; Lautenschlager et al., 2014) as imaged by micro-CT has been published.
Bone can be particularly well imaged with this technique, although care has to be taken to prevent movement during the scanning procedure. The use of X-ray contrast media (contrast-enhanced CT [CE-CT]) aids the visualization of nonmineralized tissue (Metscher, 2009a) and can allow histological-scale resolution of embryonic tissue (Metscher, 2009b).
Diffusible iodine-based contrast-enhanced computed tomography (diceCT) has become increasingly popular due to its low cost, ease of handling, and its ability to differentiate between major types of soft tissue (Gignac et al., 2016). The penguin's head is larger than that of both the dove and common buzzard. However, similarly sized soft tissue structures from other animals have been successfully imaged using diceCT (Gignac & Kley, 2014;Gignac et al., 2016;Lupon et al., 2020).
Here, our aim was to develop easily interpretable illustrations of the extracranial anatomy of the penguin head, equal in quality to those seen in human anatomy texts, to support teaching, and research and to aid veterinarians in the interpretation of diagnostic imaging and when undertaking therapeutic interventions. We concentrated on the bones and large organs, excluding the illustration of Systema Nervosa and Systema Cardiovasculare, both because of the condition in which the specimens arrived and for the sake of brevity. We have also included in the associated repository reconstructions from the micro-CT images, including files that can be 3D-printed for didactic purposes. By developing these resources for penguins, we also aimed to show the utility both of anatomic illustration as a facilitator in the understanding of anatomy and of micro-CT for resolving fine anatomic features, and thereby encourage the development of similar resources for other important research taxa.
We elected to use the head of a little penguin (Eudyptula minor) for this purpose. This animal is known in te reo Māori as kororā and we have used these common names interchangeably. We also examined the heads of both gentoo (Pygoscelis papua) and king (Aptenodytes patagonicus) penguins, concentrating on the skeletal elements. We anticipated that the highest resolution images would be obtained from the little penguin, given the smaller tissue volume and the scanner able to be used for this head, and this influenced our decision to concentrate on that species.
Although primarily the reason for selecting these species was their availability in Auckland, New Zealand, nevertheless they include both the smallest penguin, the kororā, which rarely dives below 60 m (Montague, 1985), as well as the second-largest extant penguin, the king, capable of diving to over 300 m (Culik et al., 1996). They also occupy a diverse range of habitats, from warm temperate shores to subantarctic islands (Shirihai, 2007), and represent different lineages within Spheniscidae (Baker et al., 2006;Zusi, 1975), with recent DNA evidence suggesting that Aptenodytes, perhaps along with Pygoscelis, are sister to all other extant penguins (Cole et al., 2019;Vianna et al., 2020). We aimed, therefore, to uncover novel anatomic features of Spheniscidae and perhaps differences between genera, apart from the obvious difference in size.

| Specimens
We obtained the heads of three little penguins (kororā), Eudyptula minor (Forster, 1781) (L1, L2, and L3), two gentoo penguins, Pygoscelis papua (Forster, 1781) (G1 and G2) and four king penguins (Miller, 1778;K1, K2, K3, and K4). All kororā were adults, had been found dead of natural causes in coastal areas of the Auckland Region, New Zealand and had been stored frozen before our obtaining them. The gentoo and king specimens had been bred in captivity and were resident in SEA LIFE Kelly Tarlton's Aquarium in Auckland but were of South Georgian descent and had been euthanized either because of age-related functional deterioration or, in the case of G2, failure to thrive at 7 weeks old. G2 was thus the only juvenile bird in the series. Only the heads of L1, L2, G1, and K1 were completely intact when we received them, the others having been subject to postmortem examination or, in the case of L3, cleaned by cold water maceration over a period of months. What remains of these now mostly skeletal specimens are held in the Department of Ophthalmology, Faculty

| Specimen preparation and imaging
Because stabilization of large, unfixed specimens during a long acquisition period is challenging, particularly when imaging in air, and because the heads were frozen when we received them, we imaged all specimens frozen, as previously described (De Rycke et al., 2012;Ferrare et al., 2013;Green & Gignac, 2021 After enucleation of Bulbus oculi dextri, the head of L1 was defrosted and fixed in 1% formaldehyde and 1.25% glutaraldehyde, both to improve the penetration by the contrast agent and allow visualization in air, the latter providing good contrast with a tissue while reducing the density of the specimens to X-ray. The specimen was then dehydrated using increasing concentrations of ethanol over 18 days, to remove lipid content and this reduction in tissue volume further reduces the energy required while imaging (Lupon et al., 2020).
F I G U R E 7 Eudyptula minor, left lateral view of the deeper muscles of the head (specimen L1). While Musculus cucullaris capitis is depicted as having two major divisions in this illustration, in L2 only the equivalent of the more ventral division was present (Appendix 10). Musculus depressor mandibulae appeared to have two divisions, one larger and more rostral to the other. Where anatomic nomenclature varies between Nomina Anatomica Avium (Baumel et al., 1993) and Holliday and Witmer (2007), the former has been used and the latter noted F I G U R E 8 Eudyptula minor, drawing of the deepest extracranial structures visible from the side in the head (specimen L1). Left lateral view. Where anatomic nomenclature varies between Holliday and Witmer (2007) and Nomina Anatomica Avium (Baumel et al., 1993), the latter has been used and the former noted It was then diffusion stained with alcoholic potassium triiodide (IKI) solution 0.75% for a period of 14 days before being briefly washed in 70% alcohol before rescanning using the same Bruker Skyscan 1172 and software as described above.
To confirm the anatomy obtained through imaging and to answer remaining uncertainties, following micro-CT the heads of L1 and G1 were dissected, L1 having been fixed and stained as described above while G1 was defrosted and dissected fresh. The heads of L2, G2, K3, and K4 were also dissected fresh while that of K2 had been preserved by means of immersion in a 4% formaldehyde solution for 3 months. Dissection was performed both under direct visualization and, to identify finer details, using a Lumera 700 ceiling-mounted operating microscope (Carl Zeiss AG). Photographs were taken during dissection for purposes of documentation. Soft tissue was subsequently removed from the skulls of G1, G2, and K4 by means of burial for 2 months in soil, to allow osteological examination. Two photographs were taken of the lateral aspect of the skull of L3, one straight lateral and one from an angle of 30°dorsally and 30°c audally, to create shadowing for the purpose of accurate illustration.
Two photographs were also taken of the ventral aspect, one straight ventral and one from an angle of 30°laterally and 30°caudally.
A ghost, or artifactual bony discontinuity, was observed in the transaxial planes of the image of K1, in the region of Fossa glandulae nasalis and Os jugale ( Figure 1). We initially suspected that this was due to transient vibration of the scanner during acquisition, as it is situated in a factory environment with machine shops in fairly close proximity and is not in the same class of instrument as the other scanners. However, more likely was a timing error, whereby the next scan field start is delayed momentarily, resulting in displacement. The software fuses the transaxial planes by fading out the top of the bottom and fading in the bottom of the one above. We were unable to rescan the head in the same machine because of Covid-19 related restrictions preventing our entry to the factory in which this machine was located. Therefore, to ensure that this appearance was artefactual, we reimaged this head using a Siemens SOMATOM Perspective CT scanner (Siemens Medical Solutions USA, Inc.),

(b)
F I G U R E 9 Eudyptula minor, Bulbus oculi sinistri and Musculi bulbi (specimen L1). The pupil was drawn based on our observation of a different, living kororā. When the pupil was the size drawn here it appeared slightly octagonal rather than completely round, although when dilated it appeared circular. The axial length of the globe was 14 mm and the equatorial diameter was 19 mm, as measured on micro-computed tomography (a) Left lateral view. (b) A posterior view of Bulbus oculi sinistri although the resolution afforded by this scanner was lower.
We attempted to correct the micro-CT using Amira and by the use of x/y alignment algorithms but this was not successful as the ghosting was rotational and offset from the strong bone signal.
Manually erasing the ghost areas in the affected transaxial planes using Photoshop Creative Cloud was also not successful due to thresholding. Hence no.stl file of K1 was able to be created.

| Scientific illustration
The illustrator, WCO, was unable to view any specimen in person and therefore had to draw the skulls of specimens L1, G1, and K1 using the micro-CT data as the primary source and, secondarily, the aforementioned skull photographs. As they were not imaged, the beaks of all three penguins were illustrated only using photographs of the skull, as was Epibranchiale of G1. All initial drafts were reviewed by PWH, who had access to the primary specimens and was able to make micro-CT renders from the raw data as required. PWH then made sketches of the modifications necessary to ensure accuracy, attaching relevant micro-CT renders and macro photographs as appropriate. This process was repeated until both were satisfied with the illustration produced. Following the completion of accurate illustrations of the skull of L1, the large soft tissue organs, namely the muscles, exocrine glands (glandulae exocrinae) and Bulbus oculi were then sketched onto the illustrations of the bony anatomy of the little penguin by PWH, using a combination of information gathered through diceCT and dissection. WCO then illustrated these sketches and these were further modified by PWH in the same manner as was done for the skulls.
Proper specimen preparation is essential to useful imaging. All our specimens had been frozen while K2, K3, K4, and G2 had also been partially dissected for postmortem examination. The head of L1 was also further treated as above to enhance micro-CT discrimination between soft tissue structures. This disruption of the vasculature these steps entailed precluded the use of perfusion staining of the vasculature. Further, neural tissue is known to be particularly sensitive to disruption by freezing and the removal of lipid during the processing, to which iodine appears to bind, is known to reduce contrast between myelinated nerves and lipid-poor tissues (Gignac & Kley, 2018;Gignac et al., 2016). Hence, Systema Cardiovasculare and Systema Nervosa were unable to be imaged.
Anatomic features were labeled according to the Nomina Anatomica Avium (NAA) (Baumel et al., 1993). Holliday and Witmer (2007) F I G U R E 10 Eudyptula minor, Apparatus hyobranchialis and associated structures (specimen L1), ventral view. With the exception of Urohyale, the laryngeal skeleton was clearly ossified on micro-computed tomography have made more recent suggestions regarding the labeling of the muscles of the adductor chamber and for this reason, we labeled these muscles with both. The NAA notes various subdivisions of Musculus pterygoideus have been described in different avian species. Zusi (1975) described this muscle in the kororā and found the same subdivisions as did we, although he altered the names he had given to these subdivisions in a subsequent paper describing the same muscle and the same subdivisions in hummingbirds (Aves: Trochlilidae; Zusi & Bentz, 1984). Given that we were also examining the species and that our findings were identical, we used his later nomenclature for these subdivisions.

| Micro-CT images and illustrations
Micro-CT skeletal reconstructions of transaxial scans of each frozen head (L1, G1, and K1) made using the.bmp raw data are presented in Figure 1; a stereoscopic image may be obtained by using red-green

| Skeletal elements
Basic data regarding skull dimensions is presented in Table 2. There were clear differences in size, which correlated with the size of the animal.
The rostral portion of Maxilla was unable to be imaged by micro-CT in any animal but disarticulated with ease in the juvenile gentoo penguin G2 ( Figure 12). Its tip, Rostrum maxillae, was formed by approached the latter bone more closely in Spheniscidae than in many other birds (Baumel et al., 1993). Flanking each Orbita ventrolaterally was the sigmoidal Arcus jugalis, a composite bone that is said to comprises part of Os maxillare, Os jugale and Os quadratojugalis. We found three separate skeletal elements to Arcus jugale on micro-CT of L1 but a distinction could not be made between Os quadratojugale and Os jugale in either G1 or K1; in the case of the latter, the resolution may have been too low to allow such visualization but this was not the case in G1 (Figure 13). Neither could we separately identify Os jugale from Os quadratojugale when we disarticulated G2 ( Figure 12). Processus rostralis, which articulated with Vomer ( Figure 12). The latter, a tall, thin, paired but fused bone, lay in the sagittal midline and in coronal section was slightly X-shaped.
Ventrolaterally, Cranium articulated with Os quadratum via two condyles, both of which lay in the same joint cavity (Articulatio quadrato-squamoso-otica). This columnar bone had a rostral projection, Processus orbitalis, from which several Musculi mandibulae took origin. The ventral portion of Os quadratum articulated with Arcus jugalis at Articulatio quadrato-quadratojugalis, Ossa mandibulae at the complex Articulatio quadratomandibularis ( Figure 14) and, medially,  Note that Processus jugalis ossis maxillaris (purple arrowheads), Os jugale (yellow arrowheads) and Os quadratojugale (green arrowheads) are able to be separately identified in this penguin, all three being present in section D. (g)-(j) Coronal micro-CT images through Arcus jugalis sinister of G1 at approximately equivalent locations to those in L1. Unlike L1, however, at no point was Os jugale able to be separated from Os quadratojugale in G1. Because Processus jugalis ossis maxillaris did not extend as far caudally in G1 as in L1, it is not present in section H Spheniscidae (Bock, 1960); the two bones remained 1.3 mm apart in L1 and 2 mm apart in K1 at the closest approach (imaging did not include this area in G1) and we could not detect an attachment on either dissection or diceCT.
The midline Basihyale was situated rostrally within Apparatus hyobranchialis and its two lateral processes articulated caudolaterally with the long straight Ceratobranchialia (Figure 16). Each Ceratobranchiale diverged at an angle of 15°to the midline in L1 and K1 (20°in G1) before articulating caudally with Epibranchiale, which curved dorsally as it headed further toward the caudal extremity of Caput. Articulating caudally with Basihyale was the midline Urohyale, which in L1 did not appear to be ossified on micro-CT, while Larynx and, on its caudal margin, Trachea, was situated ventral to Apparatus hyobranchialis.

| Soft tissue elements
The origin, path, insertion, and general description of each muscle that we were able to positively identify is summarized in Table 3, together with the figures in which they feature. Notable was the variation in extent and thickness of Musculus cucullaris capitis even between different individuals of the same species (Figure 17). In L1, the animal we elected to illustrate, there were two separate divisions, both taking origin on Crista temporalis, one slightly more dorsal and deeper to the other. Very soon after its origin both split into multiple small bundles radiating both caudally and ventrally, those of the more dorsal division heading more caudally and those of the more ventral division heading more ventrally and interdigitating with similar bundles of fibers arising off the more superficial Musculus constrictor colli. G2 was similar to L1 in this regard. However, L2, G1, and K4 had only one muscle belly and Musculus cucullaris capitis was much larger, even adjusting for the relative size of the animal, in G1 than in either L1 or L2, and it was larger again in K2, such that it enveloped the entire back of the head. Amira and CTVox reconstructions of the areas of soft tissue of primary concern to this paper are shown in The magnification afforded by a microscope clearly afforded a higher resolution than that obtained using micro-CT. However, unlike with micro-CT, the internal structure cannot be visualized in a nondestructive way nor can the joints with neighboring bones was slightly oval, with an anteroposterior diameter of 14 mm in L1, 21 mm in G1 mm, and 28 mm in K1 and a transequatorial diameter of 19 mm in L1, 30 mm in G1, and 40 mm in K1.

| DISCUSSION
As digital imaging becomes more widely available, studies that either incorporate or are solely reliant on such techniques are increasing.
Digital morphology, however, may not always yield exactly the same result as traditional dissection. In this paper, both traditional techniques and digital reconstructions were used to obtain data and thereby inform illustration. Two questions arose from this process. First, is digital imaging better than classical dissection when attempting to understand anatomy? Second, do digital reconstructions depict anatomy better than traditional illustrations?

| Digital imaging and reconstruction
The major advantages of micro-CT are the ability to image in situ, the creation of a permanent data set that is not subject to future degradation and the potential for manipulation of the original data set at a later date. Of particular utility in this study, given the back-andforth process of illustration, was the ability to repeatedly verify anatomic details. Volumetric analysis and measurement of spatial dimensions was also facilitated.
Using micro-CT on frozen specimens eliminated potential bony distortion that may occur with any form of cleaning, a requirement of traditional anatomic methods (Hendry, 1999) and clearly present in the disarticulated Os premaxillare of G2 ( Figure 12). It also maintained spatial orientation, ensuring accurate positioning of small  DiceCT required more specimen processing than did using frozen heads. Such preparation is known to cause tissue distortion and differential alteration of volumes (de Bournonville et al., 2019;Hedrick et al., 2018). However, it provided excellent visualization of soft tissues in situ. For instance, glandulae exocrinae were easily distinguished from muscles on diceCT due to a lesser affinity for iodine and CT Vox and Amira were both able to reconstruct subdivisions of complex large muscles, such as Musculus pterygoideus ( Figure 18) and Musculus adductor mandibulae externus (Figures 19 and 20), as well as enable accurate orientation of the origins of Musculi bulbi (Figure 21).
Thin tissues adjacent to air-filled cavities, particularly Conchae cavi nasi, were particularly well imaged with micro-CT. Many details of this area were visible even without staining, but the addition of contrast-enhanced the mucosal surfaces and the outline of Conchae. The images obtained allowed a better appreciation of this complex structure than that obtained with the operating microscope, particularly for L1 and G1 although less so in the lower resolution images used for K1, most notably with regard to the ability to appreciate the three-dimensional contours of these structures (Figure 22).

| Classical dissection
Despite their differences, in general, there was a good correlation between diceCT images and dissections and the two were complementary ( Figure 23). Further, there did not appear to be such major differences between fresh and fixed tissue as to impede informed illustration, our primary aim.
Dissection is cheap and does not require significant technology. It is thus more widely accessible than micro-CT. Further, if available, multiple specimens may be dissected to answer questions that arise during the process of illustration without the time and expense that micro-CT involves, and identify anatomic variation between animals, as was the case in regard to Musculus cucullaris capitis ( Figure 17). The resolution afforded by dissection can also be increased by using magnification ( Figure 14). In this study, PWH felt that traditional dissection more easily enabled accurate sketching of the external contours of muscles. It was also more accurate in the differentiation of the deep and smaller muscles of Apparatus hyobranchialis, as dissection using the operating microscope provided greater resolution than diceCT due to lesser density differentiation away from air-filled spaces. For instance, diceCT images could only confirm on some sections that the little penguin has two divisions of Musculus intermandibularis (Figure 24), a muscle which is singular in some birds but dual in others (Baumel et al., 1993; F I G U R E 16 Eudyptula minor, CT Vox reconstructions of Larynx and Trachea using micro-CT (specimen L1), demonstrating both the limit of resolution of skeletal elements and the ability of micro-CT to retain accurate spatial orientation of these small bones, useful when drawing Figure 10. (a) Dorsal and (b) ventral view of Apparatus hyobranchialis. Ossification of Cartilago cricoidea was only just visible with very fine adjustment of the transfer function in CTVox for bone, because of the small degree of density differentiation from the surrounding tissue, a result of its thinness and depth. In particular, the parasagittal Cartilago cricoidea did not appear completely ossified when reconstructed despite raw data and dissection showing that it was. It was also observed that the first two tracheal rings were open dorsally and did not form a complete ring. They were also fused to each other, except where they articulated with the caudal extension of Cartilago cricoidea.  McClearn & Noden, 1988) and a component part of Musculus Note: Some muscles took origin caudal to Caput and thus were in part outside the range of our study, as noted. Muscles which could not be reconstructed using automated segmentation on Amira and for which we were therefore reliant on dissection are also noted. constrictor ventralis (Holliday & Witmer, 2007). However, its dual nature was clearly apparent under the operating microscope and these two layers were easily able to be separated when dissecting L1. Automated Amira segmentation of Musculi pterylarum was also not possible and, although Musculus cucullaris capitis and Musculus constrictor colli were both identifiable on the raw micro-CT scans, it would not have been possible to manually reconstructed them using Amira without almost totally reliance on the dissected specimen. Thus, they were drawn solely from dissection. This may be a problem not confined to this study, as we note that these muscles were not identified on a digital dissection of the rock dove (Jones et al., 2019). Fat bodies were also present on dissection, most prominently in gentoo G1, in the depression above the nose, along either side of the dorsal oral mucosa and below the rostral Apparatus hyobranchialis. However, the processing of L1 removed such deposits from the diceCT images, leaving an empty void ( Figure 25). Finally, previous authors have come to contradictory conclusions regarding the patency of the external Nares gymnorhinales in Eudyptula (Pycraft, 1903;Zusi, 1975), of interest given its total occlusion and replacement by secondary external nares in the plunge-diving northern gannet Morus bassana (McDonald, 1960). This inconsistency was also unable to be resolved using micro-CT but simple irrigation of sinus antorbitalis with water showed that the Nares were patent in L2, G1, and G2.
On the other hand, dissection is a destructive technique and the same specimen cannot be dissected twice, although a photographic record can be made during the process. It also requires tissue manipulation, which can alter specimen interpretation. Dissection may be performed either on fresh tissue or formalin-fixed tissue. Fixation allows for longer-term preservation but, as with dice CT, causes more tissue alteration and is itself time-consuming. The texture and color of tissues appear different in fresh versus fixed tissue. Further, when Integumentum is removed and the underlying soft tissues exposed to air, they dry out and contract away from each other, a process that in this study began within a matter of hours and further altered the morphology. In this study, we were unsure as to how distinct the two divisions of Musculus depressor mandibulae, as described by Sosa

| Digital reconstruction
A clear advantage of digital reconstruction over traditional illustration is the avoidance of the labor-intensive process of the latter. Not only is an experienced artist required, but in all cases, the initial illustrations in this study had to be revised multiple times using both the actual specimen and micro-CT images (minimum 2 and maximum 9) before they were considered accurate. Furthermore, no artist can possibly hope to make an illustration that reflects exactly the raw digital data in the way that a computer program can. However, some human input and interpretation is required even with digital renders. For instance, although automated digital segmentation using Amira is accurate for bone, where the contrast with other tissues is extremely high, segmentation of muscles requires an operator to trace their outline to obtain any useful image, which dissection can usefully inform. An awareness of potential errors introduced by perspective is also required, particularly when viewing two-dimensional renders of three-dimensional objects (Hadden et al., 2020).

| Traditional illustration
Illustration and digital reconstruction should not be thought of mutually exclusive since the latter can inform the former, as was done in this study. Illustration possesses the ability to incorporate findings from both dissection and micro-CT renders, an advantage that traditional illustration will always have over digital reconstructions, given that the latter, unless given specific manual instruction to the contrary, can only utilize digitally acquired raw data. The advent of micro-CT can only serve to make illustration more accurate than if it F I G U R E 20 Eudyptula minor, CTVox reconstruction of the head (specimen L1), right lateral view, demonstrating the ability of diceCT to resolve the components and the orientation of the muscle fibers within each subdivision of Musculus adductor mandibulae externus dexter, which aided in the sketching of these details F I G U R E 21 Eudyptula minor, Amira 2021.2 reconstruction of the head (specimen L1), right lateral view, demonstrating the ability of diceCT to view Musculi bulbi in relation to each other. This was a particular advantage when drawing Figure 8 (especially for Musculus obliquus dorsalis and Musculus obliquus ventralis), as during the process of dissection Bulbus oculi shrank considerably and the enucleation required to reveal these muscles disinserted them, distorting their anatomy considerably. On the other hand, the more distal portions of the muscles are thinner and unable to be automatically segmented, thus Figure 9A relied on dissection.
Overall length = 90 mm were to rely on dissection alone, and thus more informative. An informed illustrator can also anticipate and compensate for changes that occur irrespective of specimen preparation and affect all forms of postmortem examination, such as the loss of volume in Bulbus oculi that occurs at death. As we hope the illustrations in this study demonstrate, a skilled artist can also, without changing the basic information presented, produce an illustration that is more aesthetically pleasing and more easily digested by the human brain than a digital reconstruction, and can emphasize notable features. This can be of particular importance to those new to the subject. Hence, just as landscape painting and portraiture has survived the advent of photography, so too we believe shall anatomic illustration.

| Future considerations
In similar future studies, consideration could be given to leaving the specimen in iodine for longer to improve staining and allow F I G U R E 22 Coronal micro-CT reconstructions using CT Vox of L1, G1, and K1 at corresponding points of Cavum nasi, demonstrating both the lack of distortion in these delicate structures that in situ imaging allows and the detail achievable, particularly in stained specimens, when imaging next to air-filled cavities where density changes are large. (a) Stained head (diceCT) of kororā (Eudyptula minor) L1. (b) Unstained head of gentoo penguin (Pygoscelis papua) G1. (c) Unstained head of king penguin (Aptenodytes patagonicus) K1. More soft tissue details were visible in G1 and L1 than in K1, which was imaged using a lower resolution scanner. DiceCT added definition, especially to the mucosal surfaces of both Os and Cavum nasi, in L1. L1 had a much more arched Cavum nasi than did G1 while that of K1 was even flatter, perhaps a necessary result of the relative elongation of Maxilla in the latter F I G U R E 23 Despite their differences, good correlation could be seen between micro-CT and anatomic dissection, even across genus boundaries, here with particular regard to Musculus pterygoideus, allowing both to be used in a complementary fashion when illustrating. Musculus rectus capitis lateralis (MRCL) was also visible in this view deeper penetration (Gignac et al., 2016). However, there would likely still be differential staining between tissues closer to external surfaces and those deeper in such large specimens and thus require a compromise, although different, in the choice of scanning parameters. Greater resolution of individual structures would have been possible were we to have dissected out and imaged each element separately, as we have done previously for Anulus ossicularis sclerae (Hadden et al., 2021), although the relationships between structures would be lost. Perfusion-based contrast enhancement using fresh specimens would be necessary were one to wish to image Systema Cardiovasculare. Finally, the illustration of a wider range of penguin species would further our understanding of the anatomic variation within Spheniscidae, as has been previously described for both extant and extinct species (Tambussi et al., 2015).

| CONCLUSIONS
Micro-CT scanning allows for detailed anatomic illustration, with frozen heads being useful for bony details and the avoidance of movement artefact, while fixation, dehydration, and staining allow for the differentiation of soft tissues and orientation of fibers within muscles. It is complementary to rather than a replacement for traditional dissection, just as digital reconstructions complement and inform illustration. Larger scanners can be used to image larger specimens, but such scanners have reduced resolving power and the longer X-ray path length also reduces resolution. Penetration of iodine is limited by tissue volume, but the resolution of deeper structures can be improved by the removal of overlying tissue. It is likely that micro-CT technology will continue to develop with time and that this will lead to improved resolution and research groups such as those that have led the development of dicect.com provide an expanding resource for others who wish to commence this type of study.
Detailed anatomic analysis of micro-CT images may also offer new insights into penguin morphology although some aspects of the anatomy, such as the patency of the nares, may always be easier to ascertain using fresh tissue; a wider range of individuals than were in this study would be required to make definitive statements in this regard. Morphological differences correlate with diet and other behaviors and therefore are of use in determining trophic habits, including those of extinct penguins that one can no longer observe F I G U R E 24 Eudyptula minor, coronal micro-computed tomography (CT) section at the level of Larynx (specimen L1), demonstrating the lesser ability of micro-CT to differentiate deeper, small soft tissue structures. Two separate divisions of Musculus intermandibularis, namely ventralis (MIV) and dorsalis (MID), were visible on the left of the skull (right of the image) but not the right, although on dissection with an operating microscope two were present throughout. Contrast between muscle and other soft tissues was better when they were adjacent to air-filled spaces, including the external surface of the head and the empty right socket (left of the image), due to increased penetration of the iodine contrast material in these areas F I G U R E 25 Fat bodies in Spheniscidae, visible on dissected specimens but removed by specimen preparation for diceCT and thus impossible to view on digital reconstructions. (a) Dorsal view of the skull of gentoo penguin (Pygoscelis papua) G1. A fat body was present in the depression above Nasus, between Calvaria and Maxilla. (b) Ventral view of Apparatus hyobranchialis and Trachea of G1. A fat body was visible ventral to the rostral part of Apparatus hyobranchialis. (c) Little penguin (Eudyptula minor) L1, coronal micro-computed tomography section. The fat body dorsal to Nasus was removed by the process of staining, leaving a "space"