By the Scientific Affairs Committee
Artificial intelligence (AI) has advanced sufficiently that it is increasingly being used as a tool throughout our society. While it can save time and intuitively provide solutions to problems ranging from every day to specialized, it also raises questions of validity and ethics.
ChatGPT, launched in November 2022 by OpenAI Inc., has quickly become the most used AI platform by both scientists and non-scientists, and has extensively validated knowledge of human anatomy. Though this knowledge needs improving with anatomical science information from diverse ethnicities, students can use it to answer specific inquiries and generate practice questions. New (yet to be validated) voice and image capabilities open up even greater educational possibilities.
AI is also being deployed in medical education through 3D anatomical model development for virtual dissections enhanced with virtual and augmented reality tools which may help solve the disparity between the supply and demand for body donation. Although AI based methods solve these and other problems such as the need for deidentification of the human donor, authentic dissection experience is invaluable, and probably AI should be supplemental to hands-on cadaveric dissection.
AI has also begun revolutionizing some biological research methods. While ChatGPT cannot yet provide useful bioinformatic analysis of large genomic, transcriptomic and proteomic datasets, it can provide functional syntax for both R and python. Also, the AI platform AlphaFold can predict the folding structure of any protein, improving the ability to design drugs against specific target molecules. When used to analyze organoids, AI provides more efficient and quantitative information on cell types and viability. AI algorithms have also been used to improve identification of abnormal peripheral-blood leukocytes, improving the accuracy, shortening the classification time and reducing the missed diagnosis rate of abnormal cells. However, AI cannot yet completely replace professional histologist input. Classifications still need to be confirmed by morphological experts, especially when encountering novel phenotypes.
Similarly, AI is facilitating high-throughput analysis of image data in various biomedical imaging disciplines. AI tools have been successfully applied to image segmentation of biomedical CT as well as on paleontological microCT images to detect bone vs. non-bone materials. Continued advances in data collection and aggregation, processing power, deep learning algorithms, and convolutional neural networks are improving the accuracy of cell detection and segmentation for instance through the estimation of structural abnormalities of retinal cone- and rod-function and improved diagnosis of neurovascular age-related macular degeneration. Additionally, AI platforms are improving scientific article searches, and their ability to summarize may aid in drafting of abstracts and conclusions preventing author tunnel vision and the refinement of writing. However, AI is currently poor at original scientific writing and should be viewed strictly as a writing aid and not a substitution for the traditional writing process.
In summary, although it brings with more ethical dilemmas than we can discuss in this short piece, AI is already improving anatomical research and medical education. While it is not yet independent of the expert-touch, it will continue to evolve with further use, forcing research and education to evolve along with it.