(1) In this study, we developed a deep learning (DL) model that can be used to predict late bladder toxicity. (2) We collected data obtained from 281 uterine cervical cancer patients who underwent definitive radiation therapy. The DL model was trained using 16 features, including patient, tumor, treatment, and dose parameters, and its performance was compared with that of a multivariable logistic regression model using the following metrics: accuracy, prediction, recall, F1-score, and area under the receiver operating characteristic curve (AUROC). In addition, permutation feature importance was calculated to interpret the DL model for each feature, and the lightweight DL model was designed to focus on the top five important features. (3) The DL model outperformed the multivariable logistic regression model on our dataset. It achieved an F1-score of 0.76 and an AUROC of 0.81, while the corresponding values for the multivariable logistic regression were 0.14 and 0.43, respectively. The DL model identified the doses for the most exposed 2 cc volume of the bladder (BD_2cc) as the most important feature, followed by BD_5cc and the ICRU bladder point. In the case of the lightweight DL model, the F-score and AUROC were 0.90 and 0.91, respectively. (4) The DL models exhibited superior performance in predicting late bladder toxicity compared with the statistical method. Through the interpretation of the model, it further emphasized its potential for improving patient outcomes and minimizing treatment-related complications with a high level of reliability.

This study aimed to investigate the performance of a deep learning-based survival-prediction model, which predicts the overall survival (OS) time of glioblastoma patients who have received surgery followed by concurrent chemoradiotherapy (CCRT). The medical records of glioblastoma patients who had received surgery and CCRT between January 2011 and December 2017 were retrospectively reviewed. Based on our inclusion criteria, 118 patients were selected and semi-randomly allocated to training and test datasets (3:1 ratio, respectively). A convolutional neural network–based deep learning model was trained with magnetic resonance imaging (MRI) data and clinical profiles to predict OS. The MRI was reconstructed by using four pulse sequences (22 slices) and nine images were selected based on the longest slice of glioblastoma by a physician for each pulse sequence. The clinical profiles consist of personal, genetic, and treatment factors. The concordance index (C-index) and integrated area under the curve (iAUC) of the time-dependent area-under-the-curve curves of each model were calculated to evaluate the performance of the survival-prediction models. The model that incorporated clinical and radiomic features showed a higher C-index (0.768 (95% confidence interval (CI): 0.759, 0.776)) and iAUC (0.790 (95% CI: 0.783, 0.797)) than the model using clinical features alone (C-index = 0.693 (95% CI: 0.685, 0.701); iAUC = 0.723 (95% CI: 0.716, 0.731)) and the model using radiomic features alone (C-index = 0.590 (95% CI: 0.579, 0.600); iAUC = 0.614 (95% CI: 0.607, 0.621)). These improvements to the C-indexes and iAUCs were validated using the 1000-times bootstrapping method; all were statistically significant (p < 0.001). This study suggests the synergistic benefits of using both clinical and radiomic parameters. Furthermore, it indicates the potential of multi-parametric deep learning models for the survival prediction of glioblastoma patients.

Background/Objectives: Large language models (LLMs) show promise in healthcare but face challenges with hallucinations, particularly in rapidly evolving fields like diabetes management. Traditional LLM updating methods are resource-intensive, necessitating new approaches for delivering reliable, current medical information. This study aimed to develop and evaluate a novel retrieval system to enhance LLM reliability in diabetes management across different languages and guidelines. Methods: We developed a dual retrieval-augmented generation (RAG) system integrating both Korean Diabetes Association and American Diabetes Association 2023 guidelines. The system employed dense retrieval with 11 embedding models (including OpenAI, Upstage, and multilingual models) and sparse retrieval using BM25 algorithm with language-specific tokenizers. Performance was evaluated across different top-k values, leading to optimized ensemble retrievers for each guideline. Results: For dense retrievers, Upstage’s Solar Embedding-1-large and OpenAI’s text-embedding-3-large showed superior performance for Korean and English guidelines, respectively. Multilingual models outperformed language-specific models in both cases. For sparse retrievers, the ko_kiwi tokenizer demonstrated superior performance for Korean text, while both ko_kiwi and porter_stemmer showed comparable effectiveness for English text. The ensemble retrievers, combining optimal dense and sparse configurations, demonstrated enhanced coverage while maintaining precision. Conclusions: This study presents an effective dual RAG system that enhances LLM reliability in diabetes management across different languages. The successful implementation with both Korean and American guidelines demonstrates the system’s cross-regional capability, laying a foundation for more trustworthy AI-assisted healthcare applications.

Gold nanoparticles (GNPs) injected in a body for dose enhancement in radiation therapy are known to form clusters. We investigated the dependence of dose enhancement on the GNP morphology using Monte-Carlo simulations and compared the model predictions with experimental data. The cluster morphology was approximated as a body-centred cubic (BCC) structure by placing GNPs at the 8 corners and the centre of a cube with an edge length of 0.22–1.03 µ m in a 4 × 4 × 4 µ m ³ water-filled phantom. We computed the dose enhancement ratio (DER) for 50 and 260 kVp photons as a function of the distance from the cube centre for 12 different cube sizes. A 10 nm-wide concentric shell shaped detector was placed up to 100 nm away from a GNP at the cube centre. For model validation, simulations based on BCC and nanoparticle random distribution (NRD) models were performed using parameters that corresponded to the experimental conditions, which measured increases in the relative biological effect due to GNPs. We employed the linear quadratic model to compute cell surviving fraction (SF) and sensitizer enhancement ratio (SER). The DER is inversely proportional to the distance to the GNPs. The largest DERs were 1.97 and 1.80 for 50 kVp and 260 kVp photons, respectively. The SF predicted by the BCC model agreed with the experimental value within 10%, up to a 5 Gy dose, while the NRD model showed a deviation larger than 10%. The SERs were 1.21 ± 0.13, 1.16 ± 0.11, and 1.08 ± 0.11 according to the experiment, BCC, and NRD models, respectively. We most accurately predicted the GNP radiosensitization effect using the BCC approximation and suggest that the BCC model is effective for use in nanoparticle dosimetry.

Purpose: Lattice radiotherapy (LRT), a type of spatially fractionated radiotherapy (SFRT), delivers high dose at specific volumes of lattice structure within the tumor to create a low valley-to-peak dose ratio (VPDR). This study aims to evaluate the feasibility of implementing SFRT using helical tomotherapy and to investigate the effects of vertex size and spacing for attaining the VPDR.Methods: A three-dimensional lattice structure with 3×3×3 vertices was designed in a cheese phantom. Vertex sizes of 0.5 cm, 1.0 cm, and 2.0 cm were assessed, with spacing from 1.0 cm to 5.0 cm. The prescribed dose was set to 20 Gy to the vertices in a single fraction. VPDR was calculated from dose profiles along lines connecting three vertices in the anterior-posterior (AP), lateral (LAT), and superior-inferior (SI) directions. The minimum, maximum, and mean dose for each vertex, as well as conformity, homogeneity and monitor unit (MU) analysis were also performed.Results: VPDR decreased significantly with increasing vertex size and spacing. While the AP and LAT directions showed similar VPDR values, the SI direction consistently exhibited lower VPDR values across all configurations. Vertex sizes of 0.5 cm, 1.0 cm, and 2.0 cm required spacing of at least 3.0 cm, 2.0 cm, and 1.0 cm, respectively, to achieve VPDR values below 0.4. The conformity indices ranged from 1.0 to 4.02, and the homogeneity indices ranged from 1.20 to 1.57 across all configurations. Additionally, the MUs increased with both vertex size and spacing.Conclusions: This study quantitatively analyzed the impact of various vertex sizes and spacings on VPDR in lattice radiotherapy using helical tomotherapy. VPDR decreased with increasing vertex size and spacing, with consistently lower values in the SI direction. These findings provide crucial insights for optimizing LRT plans. The identified relationships between the parameters and VPDR offer a foundation for developing more effective LRT protocols in helical tomotherapy, potentially improving therapeutic outcomes