Creating a Global, AI-Competent Image Library of Human Skeletal 3D Models

Research in anthropology has traditionally relied on access to physical specimens of human skeletons to study human variation through time and space. Accessing collections and collecting data are both time consuming and costly, limiting the study of human variation for researchers, and especially for students. Researchers must travel vast distances to access specimens and may only have access to those specimens for a brief period of time. Students are often even more limited because their educational quality depends on whether their institution can afford to pay thousands of dollars for skeletal models to teach students the nuances of the human skeleton. Prior to 2020, many teaching collections in the United States (and other countries with colonial histories) were unethically sourced and have rightfully been removed from curriculums, but their absence means that students cannot develop practical experience, preventing them from conducting certain types of research for their careers where a more profound expertise is required. WashU has discontinued use of its own teaching collection, part of the Terry Collection, and our project will allow students to gain the experience they need in an ethical manner, with the generous contribution of individuals who donated their remains for scientific research and learning. Providing students of all backgrounds a way to learn about anthropology will allow for more diverse viewpoints and new breakthroughs in our understanding of human evolution.

The website provides a search function that allows researchers to identify suitable bone elements using a series of specifically designed filters, create collections and conduct supervised machine learning (ML) analyses. No such portal exists at the moment, as digital human remains collections are either only accessible through traditional data sharing methods in direct conversation with the home institution, while data repositories that contain human remains 3D models only contain few specimens with limited metadata and no search/filter functions specific to the human skeleton and osteological research. Therefore, our project offers to save valuable research time for identifying and assembling suitable images for analyses. Our pilot also includes tools for supervised ML similar to those that are currently being developed by researchers around the world independently; by linking ML tools directly to the portal itself, we are creating a unique bridge between data and analytics that enables users of different skill levels to access and generate data.

Ethical Data Sharing: In collaboration with MorphoSource, our project provides an open-access platform to upload and share 3D models and their metadata. Data sharing is only slowly becoming a standard in the field of anthropology as major external grant agencies require researchers to share their data through open-access repositories. Traditionally, researchers would access collections to collect visual and other data (metric, observational) that is stored on personal hard drives or data clouds and only shared upon request. This “proprietary” model of data “ownership” is no longer ethical or feasible in the 21st century. When it is ethical to do so, our website encourages researchers to share their data, with the understanding that contributions will be acknowledged in any research products that result from the data use. Our data sharing agreements require contributors to identify the type of consent or permission by which the data was obtained. We anticipate two types of data that can be shared ethically: images of human remains that have living consent or next-of-kin consent for research and educational purposes, as well as human remains from historical or archaeological collections that have been obtained with the appropriate permissions from governmental or community agencies.

Global Scope: Human remains collections are traditionally homogenous (compared to global human variation) as they represent individuals from a single time or place. Although comparative research has been conducted, many of the existing models used as standards in human evolution and forensics, for example, are population specific and cannot easily be applied to other populations. A global data sharing platform, such as the one designed by this project, will facilitate globally comparative research, allowing the creation of population-sensitive standards and new insights into phenotypical human variation.

Beyond Anthropology: The project's website and AI tools facilitate sharing and using data among a global community of researchers who are interested in human variation through time and space, including human evolution and ancestry studies, health and illness in past and present populations, and victim identification in forensic contexts. The trajectory of this project points to a time where users, after receiving permission, can upload their bone scans and have our classification algorithms estimate features. This technology would have significant forensic and medical implications. Forensic anthropologists examine skeletal remains from criminal investigations, missing persons cases, and disaster scenarios where they need to build a profile including features like age and sex. Given the labor cost and specialized expertise required for these assessments, they are not always possible. Having a validated web accessible tool to supplement these assessments when resources are low could significantly improve identification processes.

1. Populate MorphoSource database with 3D models
2. Build out collaborations with researchers and institutions
3. Establish brand identity and publish official use/ethical guidelines
4. Implement additional site features
5. Expand automated measurement extraction and landmark placement to new bone types