Cancer is an abnormal growth with potential to invade locally and metastasize to distant organs. Accurate auto-segmentation of the tumor and surrounding normal tissues is required for radiotherapy treatment plan optimization. Recent AI-based segmentation models are generally trained on large public datasets, which lack the heterogeneity of local patient populations. While these studies advance AI-based medical image segmentation, research on local datasets is necessary to develop and integrate AI tumor segmentation models directly into hospital software for efficient and accurate oncology treatment planning and execution. This study enhances tumor segmentation using computationally efficient hybrid UNet-Transformer models on magnetic resonance imaging (MRI) datasets acquired from a local hospital under strict privacy protection. We developed a robust data pipeline for seamless DICOM extraction and preprocessing, followed by extensive image augmentation to ensure model generalization across diverse clinical settings, resulting in a total dataset of 6080 images for training. Our novel architecture integrates UNet-based convolutional neural networks with a transformer bottleneck and complementary attention modules, including efficient attention, Squeeze-and-Excitation (SE) blocks, Convolutional Block Attention Module (CBAM), and ResNeXt blocks. To accelerate convergence and reduce computational demands, we used a maximum batch size of 8 and initialized the encoder with pretrained ImageNet weights, training the model on dual NVIDIA T4 GPUs via checkpointing to overcome Kaggle's runtime limits. Quantitative evaluation on the local MRI dataset yielded a Dice similarity coefficient of 0.764 and an Intersection over Union (IoU) of 0.736, demonstrating competitive performance despite limited data and underscoring the importance of site-specific model development for clinical deployment.

In clinical practice, imaging modalities with functional characteristics, such as positron emission tomography (PET) and fractional anisotropy (FA), are often aligned with a structural reference (e.g., MRI, CT) for accurate interpretation or group analysis, necessitating multi-modal deformable image registration (DIR). However, due to the extreme heterogeneity of these modalities compared to standard structural scans, conventional unsupervised DIR methods struggle to learn reliable spatial mappings and often distort images. We find that the similarity metrics guiding these models fail to capture alignment between highly disparate modalities. To address this, we propose M2M-Reg (Multi-to-Mono Registration), a novel framework that trains multi-modal DIR models using only mono-modal similarity while preserving the established architectural paradigm for seamless integration into existing models. We also introduce GradCyCon, a regularizer that leverages M2M-Reg's cyclic training scheme to promote diffeomorphism. Furthermore, our framework naturally extends to a semi-supervised setting, integrating pre-aligned and unaligned pairs only, without requiring ground-truth transformations or segmentation masks. Experiments on the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset demonstrate that M2M-Reg achieves up to 2x higher DSC than prior methods for PET-MRI and FA-MRI registration, highlighting its effectiveness in handling highly heterogeneous multi-modal DIR. Our code is available at https://github.com/MICV-yonsei/M2M-Reg.

Introduction and ObjectivesMany women with multiple sclerosis (MS) experience neurogenic overactive bladder (NOAB) characterized by urinary frequency, urinary urgency and urgency incontinence. The objective of the study was to create machine learning (ML) models utilizing clinical and imaging data to predict NOAB treatment success stratified by treatment type. MethodsThis was a retrospective cohort study of female patients with diagnosis of NOAB and MS seen at a tertiary academic center from 2017-2022. Clinical and imaging data were extracted. Three types of NOAB treatment options evaluated included behavioral therapy, medication therapy and minimally invasive therapies. The primary outcome - treatment success was defined as > 50% reduction in urinary frequency, urinary urgency or a subjective perception of treatment success. For the construction of the logistic regression ML models, bivariate analyses were performed with backward selection of variables with p-values of < 0.10 and clinically relevant variables applied. For ML, the cohort was split into a training dataset (70%) and a test dataset (30%). Area under the curve (AUC) scores are calculated to evaluate model performance. ResultsThe 110 patients included had a mean age of patients were 59 years old (SD 14 years), with a predominantly White cohort (91.8%), post-menopausal (68.2%). Patients were stratified by NOAB treatment therapy type received with 70 patients (63.6%) at behavioral therapy, 58 (52.7%) with medication therapy and 44 (40%) with minimally invasive therapies. On MRI brain imaging, 63.6% of patients had > 20 lesions though majority were not active lesions. The lesions were mostly located within the supratentorial (94.5%), infratentorial (68.2%) and 58.2 infratentorial brain (63.8%) as well as in the deep white matter (53.4%). For MRI spine imaging, most of the lesions were in the cervical spine (71.8%) followed by thoracic spine (43.7%) and lumbar spine (6.4%).10.3%). After feature selection, the top 10 highest ranking features were used to train complimentary LASSO-regularized logistic regression (LR) and extreme gradient-boosted tree (XGB) models. The top-performing LR models for predicting response to behavioral, medication, and minimally invasive therapies yielded AUC values of 0.74, 0.76, and 0.83, respectively. ConclusionsUsing these top-ranked features, LR models achieved AUC values of 0.74-0.83 for prediction of treatment success based on individual factors. Further prospective evaluation is needed to better characterize and validate these identified associations.

BackgroundCT-based radio-biomarkers could provide non-invasive insights into tumour biology to risk-stratify patients. One of the limitations is laborious manual segmentation of regions-of-interest (ROI). We present a deep learning auto-segmentation pipeline for radiomic analysis. Patients and Methods153 patients with resected stage 2A-3B non-small cell lung cancer (NSCLCs) had tumours segmented using nnU-Net with review by two clinicians. The nnU-Net was pretrained with anatomical priors in non-cancerous lungs and finetuned on NSCLCs. Three ROIs were segmented: intra-tumoural, peri-tumoural, and whole lung. 1967 features were extracted using PyRadiomics. Feature reproducibility was tested using segmentation perturbations. Features were selected using minimum-redundancy-maximum-relevance with Random Forest-recursive feature elimination nested in 500 bootstraps. ResultsAuto-segmentation time was [~]36 seconds/series. Mean volumetric and surface Dice-Sorensen coefficient (DSC) scores were 0.84 ({+/-}0.28), and 0.79 ({+/-}0.34) respectively. DSC were significantly correlated with tumour shape (sphericity, diameter) and location (worse with chest wall adherence), but not batch effects (e.g. contrast, reconstruction kernel). 6.5% cases had missed segmentations; 6.5% required major changes. Pre-training on anatomical priors resulted in better segmentations compared to training on tumour-labels alone (p<0.001) and tumour with anatomical labels (p<0.001). Most radiomic features were not reproducible following perturbations and resampling. Adding radiomic features, however, did not significantly improve the clinical model in predicting 2-year disease-free survival: AUCs 0.67 (95%CI 0.59-0.75) vs 0.63 (95%CI 0.54-0.71) respectively (p=0.28). ConclusionOur study demonstrates that integrating auto-segmentation into radio-biomarker discovery is feasible with high efficiency and accuracy. Whilst radiomic analysis show limited reproducibility, our auto-segmentation may allow more robust radio-biomarker analysis using deep learning features.

A recently developed method enables automated measurement of the hallux valgus angle (HVA) and the first intermetatarsal angle (IMA) from weight-bearing foot radiographs. This approach employs bone segmentation to identify anatomical landmarks and provides standardized angle measurements based on established guidelines. While effective for HVA and IMA, preoperative radiograph analysis remains complex and requires additional measurements, such as the hallux interphalangeal angle (IPA), which has received limited research attention. To expand the previous method, which measured HVA and IMA, by incorporating the automatic measurement of IPA, evaluating its accuracy and clinical relevance. A preexisting database of manually labeled foot radiographs was used to train a U-Net neural network for segmenting bones and identifying landmarks necessary for IPA measurement. Of the 265 radiographs in the dataset, 161 were selected for training and 20 for validation. The U-Net neural network achieves a high mean Sørensen-Dice index (> 0.97). The remaining 84 radiographs were used to assess the reliability of automated IPA measurements against those taken manually by two orthopedic surgeons (O_A and O_B) using computer-based tools. Each measurement was repeated to assess intraobserver (O_A1 and O_A2) and interobserver (O_A2 and O_B) reliability. Agreement between automated and manual methods was evaluated using the Intraclass Correlation Coefficient (ICC), and Bland-Altman analysis identified systematic differences. Standard error of measurement (SEM) and Pearson correlation coefficients quantified precision and linearity, and measurement times were recorded to evaluate efficiency. The artificial intelligence (AI)-based system demonstrated excellent reliability, with ICC3.1 values of 0.92 (AI <i>vs</i> O_A2) and 0.88 (AI <i>vs</i> O_B), both statistically significant (<i>P</i> < 0.001). For manual measurements, ICC values were 0.95 (O_A2 <i>vs</i> O_A1) and 0.95 (O_A2 <i>vs</i> O_B), supporting both intraobserver and interobserver reliability. Bland-Altman analysis revealed minimal biases of: (1) 1.61° (AI <i>vs</i> O_A2); and (2) 2.54° (AI <i>vs</i> O_B), with clinically acceptable limits of agreement. The AI system also showed high precision, as evidenced by low SEM values: (1) 1.22° (O_A2 <i>vs</i> O_B); (2) 1.77° (AI <i>vs</i> O_A2); and (3) 2.09° (AI <i>vs</i> O_B). Furthermore, Pearson correlation coefficients confirmed strong linear relationships between automated and manual measurements, with <i>r</i> = 0.85 (AI <i>vs</i> O_A2) and <i>r</i> = 0.90 (AI <i>vs</i> O_B). The AI method significantly improved efficiency, completing all 84 measurements 8 times faster than manual methods, reducing the time required from an average 36 minutes to just 4.5 minutes. The proposed AI-assisted IPA measurement method shows strong clinical potential, effectively corresponding with manual measurements. Integrating IPA with HVA and IMA assessments provides a comprehensive tool for automated forefoot deformity analysis, supporting hallux valgus severity classification and preoperative planning, while offering substantial time savings in high-volume clinical settings.

This study compares the classification accuracy of novel transformer-based deep learning models (ViT and BEiT) on brain MRIs of gliomas and meningiomas through a feature-driven approach. Meta's Segment Anything Model was used for semi-automatic segmentation, therefore proposing a total neural network-based workflow for this classification task. ViT and BEiT models were finetuned to a publicly available brain MRI dataset. Gliomas/meningiomas cases (625/507) were used for training and 520 cases (260/260; gliomas/meningiomas) for testing. The extracted deep radiomic features from ViT and BEiT underwent normalization, dimensionality reduction based on the Pearson correlation coefficient (PCC), and feature selection using analysis of variance (ANOVA). A multi-layer perceptron (MLP) with 1 hidden layer, 100 units, rectified linear unit activation, and Adam optimizer was utilized. Hyperparameter tuning was performed via 5-fold cross-validation. The ViT model achieved the highest AUC on the validation dataset using 7 features, yielding an AUC of 0.985 and accuracy of 0.952. On the independent testing dataset, the model exhibited an AUC of 0.962 and an accuracy of 0.904. The BEiT model yielded an AUC of 0.939 and an accuracy of 0.871 on the testing dataset. This study demonstrates the effectiveness of transformer-based models, especially ViT, for glioma and meningioma classification, achieving high AUC scores and accuracy. However, the study is limited by the use of a single dataset, which may affect generalizability. Future work should focus on expanding datasets and further optimizing models to improve performance and applicability across different institutions. This study introduces a feature-driven methodology for glioma and meningioma classification, showcasing advancements in the accuracy and model robustness of transformer-based models.

BackgroundCognitive dysfunction often co-occurs with psychopathology. Advances in neuroimaging and machine learning have led to neural indicators that predict individual differences in cognition with reasonable performance. We examined whether these neural indicators explain the relationship between cognition and mental health in the UK Biobank cohort (n > 14000). MethodsUsing machine learning, we quantified the covariation between general cognition and 133 mental health indices and derived neural indicators of cognition from 72 neuroimaging phenotypes across diffusion-weighted MRI (dwMRI), resting-state functional MRI (rsMRI), and structural MRI (sMRI). With commonality analyses, we investigated how much of the cognition-mental health covariation is captured by each neural indicator and neural indicators combined within and across MRI modalities. ResultsThe predictive association between mental health and cognition was at out-of-sample r = 0.3. Neuroimaging phenotypes captured 2.1% to 25.8% of the cognition-mental health covariation. The highest proportion of variance explained by dwMRI was attributed to the number of streamlines connecting cortical regions (19.3%), by rsMRI through functional connectivity between 55 large-scale networks (25.8%), and by sMRI via the volumetric characteristics of subcortical structures (21.8%). Combining neuroimaging phenotypes within modalities improved the explanation to 25.5% for dwMRI, 29.8% for rsMRI, and 31.6% for sMRI, and combining them across all MRI modalities enhanced the explanation to 48%. ConclusionsWe present an integrated approach to derive multimodal MRI markers of cognition that can be transdiagnostically linked to psychopathology. This demonstrates that the predictive ability of neural indicators extends beyond the prediction of cognition itself, enabling us to capture the cognition-mental health covariation.

New imaging techniques appearing over the last few decades have replaced procedures that were uncomfortable, of low specificity, and prone to adverse events. While computed tomography remains useful for imaging patients with seizures in acute settings, structural magnetic resonance imaging (MRI) has become the most important imaging modality for epilepsy evaluation, with adjunctive functional imaging also increasingly well established in presurgical evaluation, including positron emission tomography (PET), single photon ictal-interictal subtraction computed tomography co-registered to MRI and functional MRI for preoperative cognitive mapping. Neuroimaging in inherited metabolic epilepsies is integral to diagnosis, monitoring, and assessment of treatment response. Neurotransmitter receptor PET and magnetic resonance spectroscopy can help delineate the pathophysiology of these disorders. Machine learning and artificial intelligence analyses based on large MRI datasets composed of healthy volunteers and people with epilepsy have been initiated to detect lesions that are not found visually, particularly focal cortical dysplasia. These methods, not yet approved for patient care, depend on careful clinical correlation and training sets that fully sample broad populations.

This study explores a multi-modal deep learning approach that integrates pre-intervention neuroimaging and clinical data to predict endovascular therapy (EVT) outcomes in acute ischemic stroke patients. To this end, consecutive stroke patients undergoing EVT were included in the study, including patients with suspected Intracranial Atherosclerosis-related Large Vessel Occlusion ICAD-LVO and other refractory occlusions. A retrospective, single-center cohort of patients with anterior circulation LVO who underwent EVT between 2017-2023 was analyzed. Refractory LVO (rLVO) defined class, comprised patients who presented any of the following: final angiographic stenosis > 50 %, unsuccessful recanalization (eTICI 0-2a) or required rescue treatments (angioplasty +/- stenting). Neuroimaging data included non-contrast CT and CTA volumes, automated vascular segmentation, and CT perfusion parameters. Clinical data included demographics, comorbidities and stroke severity. Imaging features were encoded using convolutional neural networks and fused with clinical data using a DAFT module. Data were split 80 % for training (with four-fold cross-validation) and 20 % for testing. Explainability methods were used to analyze the contribution of clinical variables and regions of interest in the images. The final sample comprised 599 patients; 481 for training the model (77, 16.0 % rLVO), and 118 for testing (16, 13.6 % rLVO). The best model predicting rLVO using just imaging achieved an AUC of 0.53 ± 0.02 and F1 of 0.19 ± 0.05 while the proposed multimodal model achieved an AUC of 0.70 ± 0.02 and F1 of 0.39 ± 0.02 in testing. Combining vascular segmentation, clinical variables, and imaging data improved prediction performance over single-source models. This approach offers an early alert to procedural complexity, potentially guiding more tailored, timely intervention strategies in the EVT workflow.

Popliteal sciatic nerve block is a widely used technique for lower limb anesthesia. However, despite ultrasound guidance, the complex anatomical structures of the popliteal fossa can present challenges, potentially leading to complications. To accurately identify the bifurcation of the sciatic nerve for nerve blockade, we propose MDEANet, a deep learning-based segmentation network designed for the precise localization of nerves, muscles, and arteries in ultrasound images of the popliteal region. MDEANet incorporates Cascaded Multi-scale Atrous Convolutions (CMAC) to enhance multi-scale feature extraction, Enhanced Spatial Attention Mechanism (ESAM) to focus on key anatomical regions, and Cross-level Feature Fusion (CLFF) to improve contextual representation. This integration markedly improves segmentation of nerves, muscles, and arteries. Experimental results demonstrate that MDEANet achieves an average Intersection over Union (IoU) of 88.60% and a Dice coefficient of 93.95% across all target structures, outperforming state-of-the-art models by 1.68% in IoU and 1.66% in Dice coefficient. Specifically, for nerve segmentation, the Dice coefficient reaches 93.31%, underscoring the effectiveness of our approach. MDEANet has the potential to provide decision-support assistance for anesthesiologists, thereby enhancing the accuracy and efficiency of ultrasound-guided nerve blockade procedures.