Enhance mammography quality to increase cancer detection by implementing continuous AI-driven feedback mechanisms, ensuring reliable, consistent, and high-quality screening by the 'Perfect', 'Good', 'Moderate', and 'Inadequate' (PGMI) criteria. To assess the impact of the AI software 'b-box^TM' on mammography quality, we conducted a comparative analysis of PGMI scores. We evaluated scores 50 days before (A) and after the software's implementation in 2021 (B), along with assessments made in the first week of August 2022 (C1) and 2023 (C2), comparing them to evaluations conducted by two readers. Except for postsurgical patients, we included all diagnostic and screening mammograms from one tertiary hospital. A total of 4577 mammograms from 1220 women (mean age: 59, range: 21-94, standard deviation: 11.18) were included. 1728 images were obtained before (A) and 2330 images after the 2021 software implementation (B), along with 269 images in 2022 (C1) and 250 images in 2023 (C2). The results indicated a significant improvement in diagnostic image quality (p < 0.01). The percentage of 'Perfect' examinations rose from 22.34% to 32.27%, while 'Inadequate' images decreased from 13.31% to 5.41% in 2021, continuing the positive trend with 4.46% and 3.20% 'inadequate' images in 2022 and 2023, respectively (p < 0.01). Using a reliable software platform to perform AI-driven quality evaluation in real-time has the potential to make lasting improvements in image quality, support radiographers' professional growth, and elevate institutional quality standards and documentation simultaneously. Question How can AI-powered quality assessment reduce inadequate mammographic quality, which is known to impact sensitivity and increase the risk of interval cancers? Findings AI implementation decreased 'inadequate' mammograms from 13.31% to 3.20% and substantially improved parenchyma visualization, with consistent subgroup trends. Clinical relevance By reducing 'inadequate' mammograms and enhancing imaging quality, AI-driven tools improve diagnostic reliability and support better outcomes in breast cancer screening.

Ultrasound imaging plays an important role in fetal growth and maternal-fetal health evaluation, but due to the complicated anatomy of the fetus and image quality fluctuation, its interpretation is quite challenging. Although deep learning include Convolution Neural Networks (CNNs) have been promising, they have largely been limited to one task or the other, such as the segmentation or detection of fetal structures, thus lacking an integrated solution that accounts for the intricate interplay between anatomical structures. To overcome these limitations, Fetal-Net-a new deep learning architecture that integrates Multi-Scale-CNNs and transformer layers-was developed. The model was trained on a large, expertly annotated set of more than 12,000 ultrasound images across different anatomical planes for effective identification of fetal structures and anomaly detection. Fetal-Net achieved excellent performance in anomaly detection, with precision (96.5%), accuracy (97.5%), and recall (97.8%) showed robustness factor against various imaging settings, making it a potent means of augmenting prenatal care through refined ultrasound image interpretation.

This work uses T1, STIR, and T2 MRI sequences of the lumbar vertebrae and BMD measurements to identify osteoporosis using deep learning. An analysis of 1350 MRI images from 50 individuals who had simultaneous BMD and MRI scans was performed. The accuracy of a custom convolution neural network for osteoporosis categorization was assessed using deep learning. T2-weighted MRIs were most diagnostic. The suggested model outperformed T1 and STIR sequences with 88.5% accuracy, 88.9% sensitivity, and 76.1% F1-score. Modern deep learning models like GoogleNet, EfficientNet-B3, ResNet50, InceptionV3, and InceptionResNetV2 were compared to its performance. These designs performed well, but our model was more sensitive and accurate. This research shows that T2-weighted MRI is the best sequence for osteoporosis diagnosis and that deep learning overcomes BMD-based approaches by reducing ionizing radiation. These results support clinical use of deep learning with MRI for safe, accurate, and quick osteoporosis diagnosis.

To assess the prevalence of motion artifacts and the factors associated with them in a cohort of suspected stroke patients, and to determine their impact on diagnostic accuracy for both AI and radiologists. This retrospective cross-sectional study included brain MRI scans of consecutive adult suspected stroke patients from a non-comprehensive Danish stroke center between January and April 2020. An expert neuroradiologist identified acute ischemic, hemorrhagic, and space-occupying lesions as references. Two blinded radiology residents rated MRI image quality and motion artifacts. The diagnostic accuracy of a CE-marked deep learning tool was compared to that of radiology reports. Multivariate analysis examined associations between patient characteristics and motion artifacts. 775 patients (68 years ± 16, 420 female) were included. Acute ischemic, hemorrhagic, and space-occupying lesions were found in 216 (27.9%), 12 (1.5%), and 20 (2.6%). Motion artifacts were present in 57 (7.4%). Increasing age (OR per decade, 1.60; 95% CI: 1.26, 2.09; p < 0.001) and limb motor symptoms (OR, 2.36; 95% CI: 1.32, 4.20; p = 0.003) were independently associated with motion artifacts in multivariate analysis. Motion artifacts significantly reduced the accuracy of detecting hemorrhage. This reduction was greater for the AI tool (from 88 to 67%; p < 0.001) than for radiology reports (from 100 to 93%; p < 0.001). Ischemic and space-occupying lesion detection was not significantly affected. Motion artifacts are common in suspected stroke patients, particularly in the elderly and patients with motor symptoms, reducing accuracy for hemorrhage detection by both AI and radiologists. Question Motion artifacts reduce the quality of MRI scans, but it is unclear which factors are associated with them and how they impact diagnostic accuracy. Findings Motion artifacts occurred in 7% of suspected stroke MRI scans, associated with higher patient age and motor symptoms, lowering hemorrhage detection by AI and radiologists. Clinical relevance Motion artifacts in stroke brain MRIs significantly reduce the diagnostic accuracy of human and AI detection of intracranial hemorrhages. Elderly patients and those with motor symptoms may benefit from a greater focus on motion artifact prevention and reduction.

To develop and validate artificial intelligence models based on contrast-enhanced CT(CECT) images of venous phase using deep learning (DL) and Radiomics approaches to predict lymphovascular invasion in gastric cancer prior to surgery. We retrospectively analyzed data from 351 gastric cancer patients, randomly splitting them into two cohorts (training cohort, n = 246; testing cohort, n = 105) in a 7:3 ratio. The tumor region of interest (ROI) was outlined on venous phase CT images as the input for the development of radiomics, 2D and 3D DL models (DL2D and DL3D). Of note, by centering the analysis on the tumor's maximum cross-section and incorporating seven adjacent 2D images, we generated stable 2.5D data to establish a multi-instance learning (MIL) model. Meanwhile, the clinical and feature-combined models which integrated traditional CT enhancement parameters (Ratio), radiomics, and MIL features were also constructed. Models' performance was evaluated by the area under the curve (AUC), confusion matrices, and detailed metrics, such as sensitivity and specificity. A nomogram based on the combined model was established and applied to clinical practice. The calibration curve was used to evaluate the consistency between the predicted LVI of each model and the actual LVI of gastric cancer, and the decision curve analysis (DCA) was used to evaluate the net benefit of each model. Among the developed models, 2.5D MIL and combined models exhibited the superior performance in comparison to the clinical model, the radiomics model, the DL2D model, and the DL3D model as evidenced by the AUC values of 0.820, 0.822, 0.748, 0.725, 0.786, and 0.711 on testing set, respectively. Additionally, the 2.5D MIL and combined models also showed good calibration for LVI prediction, and could provide a net clinical benefit when the threshold probability ranged from 0.31 to 0.98, and from 0.28 to 0.84, indicating their clinical usefulness. The MIL and combined models highlight their performance in predicting preoperative lymphovascular invasion in gastric cancer, offering valuable insights for clinicians in selecting appropriate treatment options for gastric cancer patients.

Epicardial Adipose Tissue (EAT) is a recognized risk factor for cardiovascular diseases and plays a pivotal role in the pathophysiology of Atrial Fibrillation (AF). Accurate automatic segmentation of the EAT around the Left Atrium (LA) from Magnetic Resonance Imaging (MRI) data remains challenging. While Convolutional Neural Networks excel at multi-scale feature extraction using stacked convolutions, they struggle to capture long-range self-similarity and hierarchical relationships, which are essential in medical image segmentation. In this study, we present and validate PoinUNet, a deep learning model that integrates a Poincaré embedding layer into a 3D UNet to enhance LA wall and fat segmentation from Dixon MRI data. By using hyperbolic space learning, PoinUNet captures complex LA and EAT relationships and addresses class imbalance and fat geometry challenges using a new loss function. Sixty-six participants, including forty-eight AF patients, were scanned at 1.5T. The first network identified fat regions, while the second utilized Poincaré embeddings and convolutional layers for precise segmentation, enhanced by fat fraction maps. PoinUNet achieved a Dice Similarity Coefficient of 0.87 and a Hausdorff distance of 9.42 on the test set. This performance surpasses state-of-the-art methods, providing accurate quantification of the LA wall and LA EAT.

Effective treatment for rectal cancer relies on accurate lymph node metastasis (LNM) staging. However, radiological criteria based on lymph node (LN) size, shape and texture morphology have limited diagnostic accuracy. In this work, we investigate applying a Variational Autoencoder (VAE) as a feature encoder model to replace the large pre-trained Convolutional Neural Network (CNN) used in existing approaches. The motivation for using a VAE is that the generative model aims to reconstruct the images, so it directly encodes visual features and meaningful patterns across the data. This leads to a disentangled and structured latent space which can be more interpretable than a CNN. Models are deployed on an in-house MRI dataset with 168 patients who did not undergo neo-adjuvant treatment. The post-operative pathological N stage was used as the ground truth to evaluate model predictions. Our proposed model 'VAE-MLP' achieved state-of-the-art performance on the MRI dataset, with cross-validated metrics of AUC 0.86 +/- 0.05, Sensitivity 0.79 +/- 0.06, and Specificity 0.85 +/- 0.05. Code is available at: https://github.com/benkeel/Lymph_Node_Classification_MIUA.

Accurate preoperative assessment of lymph node (LN) metastasis in rectal cancer guides treatment decisions, yet conventional MRI evaluation based on morphological criteria shows limited diagnostic performance. While some artificial intelligence models have been developed, they often operate as black boxes, lacking the interpretability needed for clinical trust. Moreover, these models typically evaluate nodes in isolation, overlooking the patient-level context. To address these limitations, we introduce LRMR, an LLM-Driven Relational Multi-node Ranking framework. This approach reframes the diagnostic task from a direct classification problem into a structured reasoning and ranking process. The LRMR framework operates in two stages. First, a multimodal large language model (LLM) analyzes a composite montage image of all LNs from a patient, generating a structured report that details ten distinct radiological features. Second, a text-based LLM performs pairwise comparisons of these reports between different patients, establishing a relative risk ranking based on the severity and number of adverse features. We evaluated our method on a retrospective cohort of 117 rectal cancer patients. LRMR achieved an area under the curve (AUC) of 0.7917 and an F1-score of 0.7200, outperforming a range of deep learning baselines, including ResNet50 (AUC 0.7708). Ablation studies confirmed the value of our two main contributions: removing the relational ranking stage or the structured prompting stage led to a significant performance drop, with AUCs falling to 0.6875 and 0.6458, respectively. Our work demonstrates that decoupling visual perception from cognitive reasoning through a two-stage LLM framework offers a powerful, interpretable, and effective new paradigm for assessing lymph node metastasis in rectal cancer.

The escalating adoption of high-resolution, large-field-of-view imagery amplifies the need for efficient compression methodologies. Conventional techniques frequently fail to preserve critical image details, while data-driven approaches exhibit limited generalizability. Implicit Neural Representations (INRs) present a promising alternative by learning continuous mappings from spatial coordinates to pixel intensities for individual images, thereby storing network weights rather than raw pixels and avoiding the generalization problem. However, INR-based compression of large images faces challenges including slow compression speed and suboptimal compression ratios. To address these limitations, we introduce COLI (Compressor for Large Images), a novel framework leveraging Neural Representations for Videos (NeRV). First, recognizing that INR-based compression constitutes a training process, we accelerate its convergence through a pretraining-finetuning paradigm, mixed-precision training, and reformulation of the sequential loss into a parallelizable objective. Second, capitalizing on INRs' transformation of image storage constraints into weight storage, we implement Hyper-Compression, a novel post-training technique to substantially enhance compression ratios while maintaining minimal output distortion. Evaluations across two medical imaging datasets demonstrate that COLI consistently achieves competitive or superior PSNR and SSIM metrics at significantly reduced bits per pixel (bpp), while accelerating NeRV training by up to 4 times.

Foundation models, pre-trained on large image datasets and capable of capturing rich feature representations, have recently shown potential for zero-shot image registration. However, their performance has mostly been tested in the context of rigid or less complex structures, such as the brain or abdominal organs, and it remains unclear whether these models can handle more challenging, deformable anatomy. Breast MRI registration is particularly difficult due to significant anatomical variation between patients, deformation caused by patient positioning, and the presence of thin and complex internal structure of fibroglandular tissue, where accurate alignment is crucial. Whether foundation model-based registration algorithms can address this level of complexity remains an open question. In this study, we provide a comprehensive evaluation of foundation model-based registration algorithms for breast MRI. We assess five pre-trained encoders, including DINO-v2, SAM, MedSAM, SSLSAM, and MedCLIP, across four key breast registration tasks that capture variations in different years and dates, sequences, modalities, and patient disease status (lesion versus no lesion). Our results show that foundation model-based algorithms such as SAM outperform traditional registration baselines for overall breast alignment, especially under large domain shifts, but struggle with capturing fine details of fibroglandular tissue. Interestingly, additional pre-training or fine-tuning on medical or breast-specific images in MedSAM and SSLSAM, does not improve registration performance and may even decrease it in some cases. Further work is needed to understand how domain-specific training influences registration and to explore targeted strategies that improve both global alignment and fine structure accuracy. We also publicly release our code at \href{https://github.com/mazurowski-lab/Foundation-based-reg}{Github}.