Abstract Quality assurance in radiotherapy traditionally relies on pre-treatment measurement-based verification using homogeneous phantoms, which may not fully account for patient-specific heterogeneities. Independent dose calculation (IDC) methods have emerged as an alternative for secondary plan verification, but their accuracy depends on proper modelling of multileaf collimator (MLC) parameters. This work focuses on the efficient and systematic tuning of these parameters. In particular, the MC module of RadCalc® (RCMC, version 7.4) was commissioned for multiple photon beam energies (6X, 6FFF, 10X, 10FFF) for two Varian TrueBeam® linear accelerators, namely TB1 and TB2. An analytical workflow was developed to determine the optimal Additional Radiation to Light Field Offset (ARLFO) value from a linear interpolation of the differences between dose measurements at different MLC-gaps and RCMC calculations, combined with weighting factors reflecting dosimetric sensitivity and clinical MLC-apertures frequencies. The method was validated on an independent clinical set of plans using the Gamma Passing Rate (GPR) evaluation (2%/2 mm local dose criterion, 10% dose threshold), and statistical significance of differences between optimal and default values of ARLFO was assessed via Wilcoxon signed-rank test. The obtained optimal ARLFO values differed between the two linacs, being positive for TB1 and negative for TB2 across all beam energies. The use of the optimized ARLFO led to a systematic increase in median GPRs: for TB1, improvements ranged from +3.8 to +8.4 percentage points (p ≤ 0.028); for TB2, increases ranged from +8.8 to +17 percentage points (p ≤ 0.0020), with no significant change observed for 10X (p = 0.31). The proposed analytical approach enables simple, efficient, and reproducible user-specific parameter tuning in Monte Carlo–based secondary dose verification and is expected to be applicable across different clinical settings.
Abstract The steady-state visual evoked potential (SSVEP) is widely used for electroencephalography (EEG) based brain computer interfaces (BCIs). In order to determine the target commands for SSVEP-BCIs, deep neural networks have been widely used. However, learning-based approaches require training data, and the upfront costs are high. This study proposes a training free signal enhancement method for SSVEP-BCIs. The proposed method enhances SSVEP components by repeatedly performing random sampling and pasting of EEG signals. Our experiments using synthesized signals and open benchmark datasets demonstrate the performance of the proposed method. The proposed method outperforms conventional methods for all signal lengths, demonstrating an improvement in information transfer rate of up to 11.30 bit/min.
Abstract Accurate segmentation of COVID-19 pneumonia lesions in chest X-ray (CXR) images is critical for disease quantification and clinical decision-making. However, existing methods suffer from blurred lesion boundaries, missed detection of tiny lesions, and limited generalization performance. To address these challenges, this paper proposes a language-guided multimodal medical image segmentation method integrating high-frequency enhancement and a dual-model ensemble. The method is developed along three dimensions: frequency-domain detail enhancement, spatial alignment optimization, and dual-model ensemble. First, a frequency-domain decomposition-based high-frequency enhancement module is designed to mitigate detail attenuation caused by encoder downsampling. Second, a two-dimensional (2D) spatial position encoding module is introduced to improve the spatial alignment between text descriptions and visual regions. Third, a dual-model ensemble architecture with differentiated frequency-domain masks is constructed to address the performance ceiling of individual models. Finally, experiments on two public datasets with different imaging modalities and data characteristics demonstrate that the proposed method consistently outperforms existing single-modal and multimodal baseline methods and achieves competitive performance comparable to state-of-the-art approaches, verifying the effectiveness and robustness of the proposed framework. On the QaTa-COV19 dataset, the Dice coefficient reaches 91.85% and the mean Intersection over Union (mIoU) reaches 84.98%. On the multi-institutional MosMedData+ dataset, the Dice coefficient reaches 78.48% and the mIoU reaches 64.66%. The proposed method provides effective technical support for clinical diagnosis and follow-up assessment of pneumonia.
Abstract Clinical metadata is widely used as auxiliary information in medical image analysis and provides important diagnostic cues for similar-case retrieval. With the development of biomedical language models, structured clinical metadata can be converted into natural-language descriptions and encoded as text representations. However, it remains unclear whether such verbalized metadata is consistently more effective than compact structured vectors when retrieval accuracy, computational efficiency, and robustness are considered together. We compare structured vector retrieval, sentence-level verbalized metadata retrieval, and phrase-level verbalized metadata retrieval using multiple biomedical text encoders. Experiments are conducted on three public medical datasets spanning lung CT (LIDC-IDRI), dermatology (PAD-UFES-20), and breast mammography (CBIS-DDSM). We report five-fold mean ± standard deviation and paired statistical tests, and further evaluate robustness under missing and incorrect metadata perturbations. Across the three datasets, structured metadata provides a strong and computationally efficient baseline, while sentence- and phrase-level verbalization produce dataset- and metric-dependent changes. The key paired Recall@5 and MV@5 comparisons do not show a consistent statistically significant improvement over structured vectors. Robustness trends also vary with the dataset and perturbation type. These findings indicate that compact structured metadata remains a practical baseline for medical case retrieval, while the benefit of verbalization depends on the metadata structure and evaluation setting.
Abstract This study investigates the effectiveness of bio-inspired barbed geometries applied to curved needles for percutaneous procedures. The geometries of insect stingers, such as honeybees and mosquitoes, have been applied to straight needles to reduce insertion forces, tissue deformation, and tissue damage during linear insertion. However, whether these benefits extend to curved needles undergoing curvilinear base manipulation remains unclear. In this study, mosquito-inspired barb geometry was adapted onto bevel-tip curved needles with bent angles of 7.5°, 15°, 22.5°, and 30°. The effect on transverse and axial insertion forces, bending moments, and needle-tip trajectories was evaluated. The regular and bio-inspired curved needles were fabricated using high-resolution 3D printing and tested in tissue-mimicking gel under curvilinear insertion. Force and moment data were recorded at the needle base, and the needle tip paths were quantified using radius of curvature analysis from video tracking. No statistically significant reductions were found in insertion forces or bending moments under curvilinear insertion. These findings suggest that, for curved needles following predefined trajectories under the tested conditions, the measured forces did not change significantly with the addition of barbed surface features under curvilinear insertion. This suggests that the tested surface modification does not substantially alter the overall mechanical response under these conditions. This work highlights that friction-reducing surface features effective in linear insertion may not directly translate to curvilinear insertion, and that optimization strategies for curved needle applications may benefit from prioritizing geometric and trajectory-based design considerations.
