yasui keisuke

To compare the geometric accuracy of two Asia‑developed AI auto‑contouring systems (RatoGuide and OncoStudio) for prostate radiotherapy planning CT, including performance in clinical edge cases. Planning CT data from 45 patients were stratified into four groups (Normal, Spacer, Seed/Spacer, GM/Spacer). Manual contours of prostate, rectum, bladder, seminal vesicles, and femoral heads served as reference. Segmentation accuracy was assessed using DSC and 95% Hausdorff distance; Wilcoxon signed‑rank and Mann-Whitney U tests were applied, with edge‑case analyses treated as exploratory. Median DSC exceeded 0.7 for all structures. RatoGuide performed better for the rectum, whereas OncoStudio performed better for the bladder, seminal vesicles, and femoral heads. Outliers were concentrated in combined implant/spacer groups. Both tools showed generally favorable geometric agreement, with structure‑dependent advantages and edge‑case outliers that warrant workflow‑aware quality assurance and future dosimetric validation.

Abstract Large language models (LLMs) have recently gained attention for their potential. However, concerns remain regarding their reliability due to limitations such as hallucinations and insufficient domain-specific knowledge. Retrieval-augmented generation (RAG) has emerged as a promising approach, enabling LLMs to reference external knowledge sources and generate accurate outputs. We aimed to clarify the potential of RAG-enhanced LLMs with Japanese input in the field of radiotherapy. This was assessed by evaluating performance on three certification examinations in Japan: the Japanese Medical Physicist Examination, the Japanese Board Examination for Radiologists, and the Japanese Board Examination for Radiation Oncologists. In this study, we constructed a RAG system named Rad-Hub, consisting of a Japanese Radiotherapy Knowledge Database (JRKD) and a retrieval framework built on Microsoft Azure. The JRKD was populated with 32 Japanese radiotherapy textbooks and clinical guidelines. We assessed its utility by inputting all multiple-choice questions from the three examinations into ChatGPT-4o, both with and without Rad-Hub, and recording the answers. They were then compared with reference answers determined by experienced medical physicists and radiation oncologists. Rad-Hub improved accuracy across all examinations. Accuracy increased from 77.0% ± 2.6% to 84.6% ± 1.5% in the Medical Physicist examination, from 74.9% ± 2.0% to 82.1% ± 1.1% in the Radiologist examination, and from 55.6% ± 4.4% to 71.7% ± 4.5% in the Radiation Oncologist examination. Performance gains ranged from 7.2% to 16.1%. These findings highlight the potential of RAG-enhanced LLMs, particularly ChatGPT-4o with Rad-Hub, for integration into radiotherapy applications, such as educational and clinical decision assistance.

ABSTRACT The patient setup using the surface‐guided radiation therapy (SGRT) system differs from conventional surface marker procedures. Owing to the abundance of three‐dimensional information, there may be operator variability in where to focus during the patient setup. This study aimed to clarify the differences between expert and novice operators in SGRT positioning for head and neck cases by tracking their eye movements, thereby providing data for developing efficient patient setup procedures. Six radiation therapists set up a simulated patient on the SGRT system while recording eye movements on the screen using the QG‐PLUS eye‐tracking system. The positioning time and number of gaze fixations on the screen were analysed, and the relationship between years of experience with SGRT, positioning time and number of gaze fixations was evaluated. No significant correlation was found between SGRT experience and positioning time ( r  = −0.67, p  = 0.15). However, more experienced radiation therapists exhibited fewer gaze fixations per positioning session ( r  = −0.81, p  < 0.05), indicating that they efficiently identified key positioning points. Additionally, experienced radiation therapists focused more intently on a specific screen during the latter half of positioning, suggesting a refined approach for final patient alignment verification. More experienced radiation therapists showed fewer gaze fixations and demonstrated increased attention to a specific screen during the latter half of the patient setup process, suggesting that eye‐tracking technology may provide useful data for standardising patient setup procedures in SGRT patient setups.