Mammography is the gold standard for the detection and diagnosis of breast cancer. This procedure can be significantly enhanced with Artificial Intelligence (AI)-based software, which assists radiologists in identifying abnormalities. However, training AI systems requires large and diverse datasets, which are often difficult to obtain due to privacy and ethical constraints. To address this issue, the paper introduces MAMmography ensemBle mOdel (MAMBO), a novel patch-based diffusion approach designed to generate full-resolution mammograms. Diffusion models have shown breakthrough results in realistic image generation, yet few studies have focused on mammograms, and none have successfully generated high-resolution outputs required to capture fine-grained features of small lesions. To achieve this, MAMBO integrates separate diffusion models to capture both local and global (image-level) contexts. The contextual information is then fed into the final patch-based model, significantly aiding the noise removal process. This thoughtful design enables MAMBO to generate highly realistic mammograms of up to 3840x3840 pixels. Importantly, this approach can be used to enhance the training of classification models and extended to anomaly detection. Experiments, both numerical and radiologist validation, assess MAMBO's capabilities in image generation, super-resolution, and anomaly detection, highlighting its potential to enhance mammography analysis for more accurate diagnoses and earlier lesion detection.

Accurate classification of second-trimester fetal ultrasound images remains challenging due to low image quality, high intra-class variability, and significant class imbalance. In this work, we introduce a simple yet powerful, biologically inspired deep learning ensemble framework that-unlike prior studies focused on only a handful of anatomical targets-simultaneously distinguishes 16 fetal structures. Drawing on the hierarchical, modular organization of biological vision systems, our model stacks two complementary branches (a "shallow" path for coarse, low-resolution cues and a "detailed" path for fine, high-resolution features), concatenating their outputs for final prediction. To our knowledge, no existing method has addressed such a large number of classes with a comparably lightweight architecture. We trained and evaluated on 5,298 routinely acquired clinical images (annotated by three experts and reconciled via Dawid-Skene), reflecting real-world noise and variability rather than a "cleaned" dataset. Despite this complexity, our ensemble (EfficientNet-B0 + EfficientNet-B6 with LDAM-Focal loss) identifies 90% of organs with accuracy > 0.75 and 75% of organs with accuracy > 0.85-performance competitive with more elaborate models applied to far fewer categories. These results demonstrate that biologically inspired modular stacking can yield robust, scalable fetal anatomy recognition in challenging clinical settings.

Differentiating major depressive disorder (MDD) from bipolar disorder (BD) remains a significant clinical challenge, as both disorders exhibit overlapping symptoms but require distinct treatment approaches. Advances in voxel-based morphometry and surface-based morphometry have facilitated the identification of structural brain abnormalities that may serve as diagnostic biomarkers. This study aimed to explore the relationships between brain morphological features, such as grey matter volume (GMV) and cortical thickness (CT), and demographic and clinical variables in patients with MDD and BD and healthy controls (HC) using multivariate analysis methods. A total of 263 participants, including 120 HC, 95 patients with MDD and 48 patients with BD, underwent T1-weighted MRI. GMV and CT were computed for standardised brain regions, followed by multivariate partial least squares (PLS) regression to assess associations with demographic and diagnostic variables. Reductions in frontotemporal CT were observed in MDD and BD compared with HC, but distinct trends between BD and MDD were also detected for the CT of selective temporal, frontal and parietal regions. Differential patterns in cerebellar GMV were also identified, with lobule CI larger in MDD and lobule CII larger in BD. Additionally, BD showed the same trend as ageing concerning reductions in CT and posterior cerebellar and striatal GMV. Depression severity showed a transdiagnostic link with reduced frontotemporal CT. This study highlights shared and distinct structural brain alterations in MDD and BD, emphasising the potential of neuroimaging biomarkers to enhance diagnostic accuracy. Accelerated cortical thinning and differential cerebellar changes in BD may serve as targets for future research and clinical interventions. Our findings underscore the value of objective neuroimaging markers in increasing the precision of mood disorder diagnoses, improving treatment outcomes.

This study aimed to differentiate between benign and malignant gallbladder polyps preoperatively by developing a prediction model integrating preoperative transabdominal ultrasound and clinical features using machine-learning algorithms. A retrospective analysis was conducted on clinical and ultrasound data from 1,050 patients at 2 centers who underwent cholecystectomy for gallbladder polyps. Six machine-learning algorithms were used to develop preoperative models for predicting benign and malignant gallbladder polyps. Internal and external test cohorts evaluated model performance. The Shapley Additive Explanations algorithm was used to understand feature importance. The main study cohort included 660 patients with benign polyps and 285 patients with malignant polyps, randomly divided into a 3:1 stratified training and internal test cohorts. The external test cohorts consisted of 73 benign and 32 malignant polyps. In the training cohort, the Shapley Additive Explanations algorithm, on the basis of variables selected by Least Absolute Shrinkage and Selection Operator regression and multivariate logistic regression, further identified 6 key predictive factors: polyp size, age, fibrinogen, carbohydrate antigen 19-9, presence of stones, and cholinesterase. Using these factors, 6 predictive models were developed. The random forest model outperformed others, with an area under the curve of 0.963, 0.940, and 0.958 in the training, internal, and external test cohorts, respectively. Compared with previous studies, the random forest model demonstrated excellent clinical utility and predictive performance. In addition, the Shapley Additive Explanations algorithm was used to visualize feature importance, and an online calculation platform was developed. The random forest model, combining preoperative ultrasound and clinical features, accurately predicts benign and malignant gallbladder polyps, offering valuable guidance for clinical decision-making.

This study aimed to develop a novel deep learning model, U₂-Attention-Net (U₂A-Net), for precise segmentation of parotid glands on radiotherapy localization CT images. CT images from 79 patients with head and neck cancer were selected, on which the label maps were delineated by relevant practitioners to construct a dataset. The dataset was divided into the training set (n = 60), validation set (n = 6), and test set (n = 13), with the training set augmented. U₂A-Net, divided into U₂A-Net V₁ (sSE) and U₂A-Net V₂ (cSE) based on different attention mechanisms, was evaluated for parotid gland segmentation based on the DL loss function with U-Net, Attention U-Net, DeepLabV3+, and TransUNet as comparision models. Segmentation was also performed using GDL and GD-BCEL loss functions. Model performance was evaluated using DSC, JSC, PPV, SE, HD, RVD, and VOE metrics. The quantitative results revealed that U₂A-Net based on DL outperformed the comparative models. While U₂A-Net V₁ had the highest PPV, U₂A-Net V₂ demonstrated the best quantitative results in other metrics. Qualitative results showed that U₂A-Net's segmentation closely matched expert delineations, reducing oversegmentation and undersegmentation, with U₂A-Net V₂ being more effective. In comparing loss functions, U₂A-Net V₁ using GD-BCEL and U₂A-Net V₂ using DL performed best. The U₂A-Net model significantly improved parotid gland segmentation on radiotherapy localization CT images. The cSE attention mechanism showed advantages with DL, while sSE performed better with GD-BCEL.

Phenotypic information for cancer research is embedded in unstructured electronic health records (EHR), requiring effort to extract. Deep learning models can automate this but face scalability issues due to privacy concerns. We evaluated techniques for applying a teacher-student framework to extract longitudinal clinical outcomes from EHRs. We focused on the challenging task of ascertaining two cancer outcomes-overall response and progression according to Response Evaluation Criteria in Solid Tumors (RECIST)-from free-text radiology reports. Teacher models with hierarchical Transformer architecture were trained on data from Dana-Farber Cancer Institute (DFCI). These models labeled public datasets (MIMIC-IV, Wiki-text) and GPT-4-generated synthetic data. "Student" models were then trained to mimic the teachers' predictions. DFCI "teacher" models achieved high performance, and student models trained on MIMIC-IV data showed comparable results, demonstrating effective knowledge transfer. However, student models trained on Wiki-text and synthetic data performed worse, emphasizing the need for in-domain public datasets for model distillation.

Speed-of-sound (SoS) is a biomechanical characteristic of tissue, and its imaging can provide a promising biomarker for diagnosis. Reconstructing SoS images from ultrasound acquisitions can be cast as a limited-angle computed-tomography problem, with variational networks being a promising model-based deep learning solution. Some acquired data frames may, however, get corrupted by noise due to, e.g., motion, lack of contact, and acoustic shadows, which in turn negatively affects the resulting SoS reconstructions. We propose to use the uncertainty in SoS reconstructions to attribute trust to each individual acquired frame. Given multiple acquisitions, we then use an uncertainty-based automatic selection among these retrospectively, to improve diagnostic decisions. We investigate uncertainty estimation based on Monte Carlo Dropout and Bayesian Variational Inference. We assess our automatic frame selection method for differential diagnosis of breast cancer, distinguishing between benign fibroadenoma and malignant carcinoma. We evaluate 21 lesions classified as BI-RADS 4, which represents suspicious cases for probable malignancy. The most trustworthy frame among four acquisitions of each lesion was identified using uncertainty-based criteria. Selecting a frame informed by uncertainty achieved an area under curve of 76% and 80% for Monte Carlo Dropout and Bayesian Variational Inference, respectively, superior to any uncertainty-uninformed baselines with the best one achieving 64%. A novel use of uncertainty estimation is proposed for selecting one of multiple data acquisitions for further processing and decision making.

Glioma molecular characterization is essential for risk stratification and treatment planning. Noninvasive imaging biomarkers such as apparent diffusion coefficient (ADC) values have shown potential for predicting glioma genotypes. However, manual segmentation of gliomas is time-consuming and operator-dependent. To address this limitation, we aimed to establish a single-sequence-derived automatic ADC extraction pipeline using T2-weighted imaging to support glioma isocitrate dehydrogenase (IDH) genotyping. Glioma volumes from a hospital data set (University College London Hospitals; n=247) were manually segmented on T2-weighted MRI scans using ITK-Snap Toolbox and co-registered to ADC maps sequences using the FMRIB Linear Image Registration Tool in FSL, followed by ADC histogram extraction (Python). Separately, a nnUNet deep learning algorithm was trained to segment glioma volumes using T2w only from BraTS 2021 data (n=500, 80% training, 5% validation and 15% test split). nnUnet was then applied to the University College London Hospitals (UCLH) data for segmentation and ADC read-outs. Univariable logistic regression was used to test the performance manual and nnUNet derived ADC metrics for IDH status prediction. Statistical equivalence was tested (paired two-sided t-test). nnUnet segmentation achieved a median Dice of 0.85 on BraTS data, and 0.83 on UCLH data. For the best performing metric (rADCmean) the area under the receiver operating characteristic curve (AUC) for differentiating IDH-mutant from IDHwildtype gliomas was 0.82 (95% CI: 0.78-0.88), compared to the manual segmentation AUC 0.84 (95% CI: 0.77-0.89). For all ADC metrics, manually and nnUNet extracted ADC were statistically equivalent (p<0.01). nnUNet identified one area of glioma infiltration missed by human observers. In 0.8% gliomas, nnUnet missed glioma components. In 6% of cases, over-segmentation of brain remote from the tumor occurred (e.g. temporal poles). The T2w trained nnUnet algorithm achieved ADC readouts for IDH genotyping with a performance statistically equivalent to human observers. This approach could support rapid ADC based identification of glioblastoma at an early disease stage, even with limited input data. AUC = Area under the receiver operating characteristic curve, BraTS = The brain tumor segmentation challenge held by MICCAI, Dice = Dice Similarity Coefficient, IDH = Isocitrate dehydrogenase, mGBM = Molecular glioblastoma, ADCmin = Fifth ADC histogram percentile, ADCmean = Mean ADC value, ADCNAWM = ADC in the contralateral centrum semiovale normal white matter, rADCmin = Normalized ADCmin, VOI rADCmean = Normalized ADCmean.

Anomaly Detection (AD) and Anomaly Localization (AL) are crucial in fields that demand high reliability, such as medical imaging and industrial monitoring. However, current AD and AL approaches are often susceptible to adversarial attacks due to limitations in training data, which typically include only normal, unlabeled samples. This study introduces PatchGuard, an adversarially robust AD and AL method that incorporates pseudo anomalies with localization masks within a Vision Transformer (ViT)-based architecture to address these vulnerabilities. We begin by examining the essential properties of pseudo anomalies, and follow it by providing theoretical insights into the attention mechanisms required to enhance the adversarial robustness of AD and AL systems. We then present our approach, which leverages Foreground-Aware Pseudo-Anomalies to overcome the deficiencies of previous anomaly-aware methods. Our method incorporates these crafted pseudo-anomaly samples into a ViT-based framework, with adversarial training guided by a novel loss function designed to improve model robustness, as supported by our theoretical analysis. Experimental results on well-established industrial and medical datasets demonstrate that PatchGuard significantly outperforms previous methods in adversarial settings, achieving performance gains of $53.2\%$ in AD and $68.5\%$ in AL, while also maintaining competitive accuracy in non-adversarial settings. The code repository is available at https://github.com/rohban-lab/PatchGuard .

The Radiological Society of North America (RSNA) Lumbar Degenerative Imaging Spine Classification (LumbarDISC) dataset is the largest publicly available dataset of adult MRI lumbar spine examinations annotated for degenerative changes. The dataset includes 2,697 patients with a total of 8,593 image series from 8 institutions across 6 countries and 5 continents. The dataset is available for free for non-commercial use via Kaggle and RSNA Medical Imaging Resource of AI (MIRA). The dataset was created for the RSNA 2024 Lumbar Spine Degenerative Classification competition where competitors developed deep learning models to grade degenerative changes in the lumbar spine. The degree of spinal canal, subarticular recess, and neural foraminal stenosis was graded at each intervertebral disc level in the lumbar spine. The images were annotated by expert volunteer neuroradiologists and musculoskeletal radiologists from the RSNA, American Society of Neuroradiology, and the American Society of Spine Radiology. This dataset aims to facilitate research and development in machine learning and lumbar spine imaging to lead to improved patient care and clinical efficiency.