<< Back to the abstract archive
Automating Data Extraction from Plastic Surgery Operation Notes with Small Artificial Intelligence Models
Christopher A. Slater, BS, Min-Jeong Cho, MD
The Ohio State University College of Medicine
2024-01-15
Presenter: Christopher Slater
Affidavit:
I certify that the material proposed for presentation in this abstract has not been published in any scientific journal or previously presented at a major meeting.
Director Name: Min-Jeong Cho
Author Category: Medical Student
Presentation Category: Basic Science Research
Abstract Category: General Reconstruction
Artificial intelligence (AI) models can drastically increase the speed and accuracy of plastic surgery research by automating the chart review process. This is especially useful for extracting information from surgical operation notes, which is an important but time-consuming process. Due to the hardware required, standard AI models require large third-party servers to operate. However, doing analysis on outside servers poses significant security vulnerabilities. Consequently, patient health information (PHI) must be de-identified before being exported to outside servers which limits the use of AI data extraction from the medical record. Using new smaller AI models, such as Mistral 7B, it is possible to run AI models on consumer grade, local computers. This removes the security vulnerabilities of third-party servers and allows for direct analysis of PHI. We propose a more accessible and scalable workflow for plastic surgery research to run AI models on local computers with increased security allowing for direct PHI analysis. In addition, all the software used is open access and the hardware needed can be assembled for less than $1,000. These systems can also be scaled up to allow for even larger models and training of new models specific to plastic surgery. Now that the workflow has been established, the next steps will be analyzing the accuracy and speed of small AI models to extract information from tissue expander surgical operation notes.
