RESUMEN
Purpose@#We aimed to evaluate the time and cost of developing prompts using large language model (LLM), tailored to extract clinical factors in breast cancer patients and their accuracy. @*Materials and Methods@#We collected data from reports of surgical pathology and ultrasound from breast cancer patients who underwent radiotherapy from 2020 to 2022. We extracted the information using the Generative Pre-trained Transformer (GPT) for Sheets and Docs extension plugin and termed this the “LLM” method. The time and cost of developing the prompts with LLM methods were assessed and compared with those spent on collecting information with “full manual” and “LLM-assisted manual” methods. To assess accuracy, 340 patients were randomly selected, and the extracted information by LLM method were compared with those collected by “full manual” method. @*Results@#Data from 2,931 patients were collected. We developed 12 prompts for Extract function and 12 for Format function to extract and standardize the information. The overall accuracy was 87.7%. For lymphovascular invasion, it was 98.2%. Developing and processing the prompts took 3.5 hours and 15 minutes, respectively. Utilizing the ChatGPT application programming interface cost US $65.8 and when factoring in the estimated wage, the total cost was US $95.4. In an estimated comparison, “LLM-assisted manual” and “LLM” methods were time- and cost-efficient compared to the “full manual” method. @*Conclusion@#Developing and facilitating prompts for LLM to derive clinical factors was efficient to extract crucial information from huge medical records. This study demonstrated the potential of the application of natural language processing using LLM model in breast cancer patients. Prompts from the current study can be re-used for other research to collect clinical information.
RESUMEN
Purpose@#The safety of online contouring and planning for adaptive radiotherapy is unknown. This study aimed to evaluate the dosimetric difference of the organ-at-risk (OAR) according to the extent of contouring in stereotactic magnetic resonance image-guided adaptive RT (SMART) for pancreatic cancer. @*Materials and Methods@#We reviewed the treatment plan data used for SMART in patients with pancreatic cancer. For the online contouring and planning, OARs within 2 cm from the planning target volume (PTV) in the craniocaudal direction were re-controlled daily at the attending physician's discretion. The entire OARs were re-contoured retrospectively for data analysis. We termed the two contouring methods the Rough OAR and the Full OAR, respectively. The proportion of dose constraint violation and other dosimetric parameters was analyzed. @*Results@#Nineteen patients with 94 fractions of SMART were included in the analysis. The dose constraint was violated in 10.6% and 43.6% of the fractions in Rough OAR and Full OAR methods, respectively (p = 0.075). Patients with a large tumor, a short distance from gross tumor volume (GTV) to OAR, and a tumor in the body or tail were associated with more occult dose constraint violations—large tumor (p = 0.027), short distance from GTV to OAR (p = 0.061), tumor in body or tail (p = 0.054). No dose constraint violation occurred outside 2 cm from the PTV. @*Conclusion@#More occult dose constraint violations can be found by the Full OAR method in patients with pancreatic cancer with some clinical factors in the online re-planning for SMART. Re-contouring all the OARs would be helpful to detect occult dose constraint violations in SMART planning. Since the dosimetric profile of SMART cannot be represented by a single fraction, patient selection for the Full OAR method should be weighted between the clinical usefulness and the time and workforce required.