In the years that I have been doing this work, I have seen the way visualizations are handled in science and education change. Partly because there is more and more hardware and software to create things faster, easier, and bigger, partly because we live in an increasingly visual world. In the medical industry, for example, you now see a tendency to use scans to prepare for surgeries. In science, scans are used, for instance, to visualize the structure of a specific organ.
For such applications, scans are, of course, perfect. They reflect a situation of that moment, that organ, that condition. But you also see scans being used for education. For specialized medical training (think gynecology, cardiology, and the like) I see no problem with that, you want students to be able to distinguish the variation within anatomical structures, but for regular educational programs, in my view, it is really more a matter of showing off and, funny enough, a lack of knowledge and skill, than anything else.
Because what do you want to teach students in regular, basic educational programs? You want to give them an insight into comparative anatomy, why different organs look the way they do, how they developed from the original germ layers, and you want to ensure that they have a genuine understanding of this. So that students can reason how, for example, an animal they do not yet know is anatomically structured. And how they can explain deviations from what they expect to see. And in medical programs, you want them to understand the evolutionary history of human anatomy.
You don’t do that with scans. Scans are the raw material, also with animals. There are too many individual specific peculiarities to efficiently teach students a good understanding of overall anatomy or taxonomy for that matter. You want to teach them the generalities instead. Throughout the centuries, various methods have been used for this. Illustrations, of course, but also physical models made of wax.
figure 5 of Takaishi et al., 2014. In vivo cancer imaging using liposomal iodinated nanoparticles. Mice are inoculated with cancer cells 3 weeks prior to imaging, allowing xenograft tumors to grow. Imaging is performed immediately after the injection of liposomal iodinated nanoparticles (early phase) and 3–7 days later (delayed phase). The early-phase scan is used to measure tumor fractional blood volume, while the delayed-phase scan is used to measure tumor vessel permeability via the enhanced permeability and retention effect.
The anatomical 3D models I make are the modern equivalent of that. But you can really do so much more with them. By linking such an anatomical model to specific data, you can in fact create a digital twin with it. Now, digital twin is obviously a buzzword, but just imagine what you can do with it. Instead of a static model that you can only view from all sides, take apart, and compare, you could show the entire development of such an animal over time, connect and compare it with the development of other animals, show how their circulatory system works, and much more.
A rendition of a rat model I am (still) working on. The muscles are removed, so is the connective tissue and addipose tissue. The model can be used for education because it is generalized.
Unfortunately, however, there is still hardly any attention given to these kinds of possibilities. Even though it would make biology education so much more interesting, research more inclusive, and I am sure it would also lead to further developments for the benefit of humans and animals that we currently have no idea about.
It is a paradoxical situation we find ourselves in today. We are stumbling over advanced hardware and the latest software systems; the technological infrastructure is practically perfect. We have the computing power to render billions of polygons in real-time and virtual reality headsets that can completely suck us into another world. But where is the content? It is as if we built a state-of-the-art printing press, only to print coloring books with it. We have the tools, but we fill them with empty shells, flashy but contentless animations, or with unprocessed, raw scan data under the guise of ‘innovation’. What is lacking is the depth of content and the vision to truly utilize that technology for fundamental understanding. We invest massively in the packaging, but forget the product.
Same anatomical model of Rattus norvegicus. Now you see part of the nervous system and some muscles.
And of course, modeling the way I do takes more manpower. Because, as has actually never changed, such models can only be made if the creator understands what he or she is making. With every organ, with every object within a model, decisions must be made so everything is both accurate and technically sound. And so that you can also use the model for multiple purposes.
It is often casually claimed that AI will easily take over this kind of modeling work. But fundamentally, AI can never be the way forward in this. Because AI works on the basis of statistics and probability; it recognizes patterns in existing data, but it understands nothing. AI does not know why a specific muscle attaches at a certain location, what evolutionary pressures led to that, or how that structure functions mechanically in relation to the entire skeleton.
A generative AI synthesizes what superficially looks plausible. But anatomy is not a matter of ‘what likely looks right’. It is a logical, biological sequence of evolution, embryonic development, and biomechanics. If you have educational models generated by an algorithm susceptible to hallucinations, you replace true understanding with an illusion. We want students to be able to reason, so we need models built with explicit intentionality, didactic insight, and the right expertise. The human translation, separating the noise from the core, is something an AI simply cannot do.
The anatomical model of Rattus norvegicus, now with a majority of the muscles and some addipose tissue.
References (for the anatomical 3d model of Rattus norvegicus that is)
[zotpress collection=”UBNCCL5X”]
