While numerous architectures for medical image segmentation have been proposed, achieving competitive performance with state-of-the-art models networks such as nnUNet, still leave room for further innovation. In this work, we introduce nnUZoo, an open source benchmarking framework built upon nnUNet, which incorporates various deep learning architectures, including CNNs, Transformers, and Mamba-based models. Using this framework, we provide a fair comparison to demystify performance claims across different medical image segmentation tasks. Additionally, in an effort to enrich the benchmarking, we explored five new architectures based on Mamba and Transformers, collectively named X2Net, and integrated them into nnUZoo for further evaluation. The proposed models combine the features of conventional U2Net, nnUNet, CNN, Transformer, and Mamba layers and architectures, called X2Net (UNETR2Net (UNETR), SwT2Net (SwinTransformer), SS2D2Net (SwinUMamba), Alt1DM2Net (LightUMamba), and MambaND2Net (MambaND)). We extensively evaluate the performance of different models on six diverse medical image segmentation datasets, including microscopy, ultrasound, CT, MRI, and PET, covering various body parts, organs, and labels. We compare their performance, in terms of dice score and computational efficiency, against their baseline models, U2Net, and nnUNet. CNN models like nnUNet and U2Net demonstrated both speed and accuracy, making them effective choices for medical image segmentation tasks. Transformer-based models, while promising for certain imaging modalities, exhibited high computational costs. Proposed Mamba-based X2Net architecture (SS2D2Net) achieved competitive accuracy with no significantly difference from nnUNet and U2Net, while using fewer parameters. However, they required significantly longer training time, highlighting a trade-off between model efficiency and computational cost.

We present a probabilistic framework for modeling structured spatiotemporal dynamics from sparse observations, focusing on cardiac motion. Our approach integrates neural ordinary differential equations (NODEs), graph neural networks (GNNs), and neural processes into a unified model that captures uncertainty, temporal continuity, and anatomical structure. We represent dynamic systems as spatiotemporal multiplex graphs and model their latent trajectories using a GNN-parameterized vector field. Given the sparse context observations at node and edge levels, the model infers a distribution over latent initial states and control variables, enabling both interpolation and extrapolation of trajectories. We validate the method on three synthetic dynamical systems (coupled pendulum, Lorenz attractor, and Kuramoto oscillators) and two real-world cardiac imaging datasets - ACDC (N=150) and UK Biobank (N=526) - demonstrating accurate reconstruction, extrapolation, and disease classification capabilities. The model accurately reconstructs trajectories and extrapolates future cardiac cycles from a single observed cycle. It achieves state-of-the-art results on the ACDC classification task (up to 99% accuracy), and detects atrial fibrillation in UK Biobank subjects with competitive performance (up to 67% accuracy). This work introduces a flexible approach for analyzing cardiac motion and offers a foundation for graph-based learning in structured biomedical spatiotemporal time-series data.

We produce a clean and well-characterised catalogue of objects within 100\,pc of the Sun from the \G\ Early Data Release 3. We characterise the catalogue through comparisons to the full data release, external catalogues, and simulations. We carry out a first analysis of the science that is possible with this sample to demonstrate its potential and best practices for its use. The selection of objects within 100\,pc from the full catalogue used selected training sets, machine-learning procedures, astrometric quantities, and solution quality indicators to determine a probability that the astrometric solution is reliable. The training set construction exploited the astrometric data, quality flags, and external photometry. For all candidates we calculated distance posterior probability densities using Bayesian procedures and mock catalogues to define priors. Any object with reliable astrometry and a non-zero probability of being within 100\,pc is included in the catalogue. We have produced a catalogue of \NFINAL\ objects that we estimate contains at least 92\% of stars of stellar type M9 within 100\,pc of the Sun. We estimate that 9\% of the stars in this catalogue probably lie outside 100\,pc, but when the distance probability function is used, a correct treatment of this contamination is possible. We produced luminosity functions with a high signal-to-noise ratio for the main-sequence stars, giants, and white dwarfs. We examined in detail the Hyades cluster, the white dwarf population, and wide-binary systems and produced candidate lists for all three samples. We detected local manifestations of several streams, superclusters, and halo objects, in which we identified 12 members of \G\ Enceladus. We present the first direct parallaxes of five objects in multiple systems within 10\,pc of the Sun.

This pilot study compares per-lesion radiomics features of [68Ga]-DOTA FAPI-46 and [18F]-FDG PET/CT in non-small cell lung cancer (NSCLC) to explore complementary insights into intratumoral heterogeneity beyond conventional SUV metrics, aiming to enhance lesion characterization and clinical decision-making. A total of 28 PET/CT scans (14 [18F]-FDG and 14 [68Ga]-DOTA FAPI-46) were acquired for the initial staging of biopsy-confirmed NSCLC. A total of 81 co-localized lesions (lung: 21, mediastinal lymph nodes: 42, bone: 18) were segmented, with radiomics features extracted via PyRadiomics after IBSI-compliant preprocessing. Paired per-lesion comparisons used t-tests or Wilcoxon signed-rank tests with Benjamini-Hochberg FDR correction. Significant differences (adjusted P < 0.05) were identified across intensity, texture, and shape features. In lung lesions, FAPI showed lower first-order metrics but higher variance and GLCM contrast, suggesting stromal heterogeneity. Mediastinal lymph nodes had fewer differences, with FAPI exhibiting lower run percentage (GLRLM: -0.298, -6.196, P=1.45E-08). Bone lesions showed extensive variations, including reduced FAPI entropy (e.g., Entropy: -1.743, -5.798, P=6.95E-08). Feature overlaps highlighted complementary stromal (FAPI) and metabolic (FDG) insights. This pilot study demonstrates that per-lesion radiomics can capture complementary biological information from FAPI and FDG PET in NSCLC, highlighting intratumoral heterogeneity and stromal activity not fully appreciated by conventional SUV-based metrics.

This paper presents FlowCyt, the first comprehensive benchmark for multi-class single-cell classification in flow cytometry data. The dataset comprises bone marrow samples from 30 patients, with each cell characterized by twelve markers. Ground truth labels identify five hematological cell types: T lymphocytes, B lymphocytes, Monocytes, Mast cells, and Hematopoietic Stem/Progenitor Cells (HSPCs). Experiments utilize supervised inductive learning and semi-supervised transductive learning on up to 1 million cells per patient. Baseline methods include Gaussian Mixture Models, XGBoost, Random Forests, Deep Neural Networks, and Graph Neural Networks (GNNs). GNNs demonstrate superior performance by exploiting spatial relationships in graph-encoded data. The benchmark allows standardized evaluation of clinically relevant classification tasks, along with exploratory analyses to gain insights into hematological cell phenotypes. This represents the first public flow cytometry benchmark with a richly annotated, heterogeneous dataset. It will empower the development and rigorous assessment of novel methodologies for single-cell analysis.

In the complex landscape of hematologic samples such as peripheral blood or bone marrow derived from flow cytometry (FC) data, cell-level prediction presents profound challenges. This work explores injecting hierarchical prior knowledge into graph neural networks (GNNs) for single-cell multi-class classification of tabular cellular data. By representing the data as graphs and encoding hierarchical relationships between classes, we propose our hierarchical plug-in method to be applied to several GNN models, namely, FCHC-GNN, and effectively designed to capture neighborhood information crucial for single-cell FC domain. Extensive experiments on our cohort of 19 distinct patients, demonstrate that incorporating hierarchical biological constraints boosts performance significantly across multiple metrics compared to baseline GNNs without such priors. The proposed approach highlights the importance of structured inductive biases for gaining improved generalization in complex biological prediction tasks.

In the realm of hematologic cell populations classification, the intricate patterns within flow cytometry data necessitate advanced analytical tools. This paper presents 'HemaGraph', a novel framework based on Graph Attention Networks (GATs) for single-cell multi-class classification of hematological cells from flow cytometry data. Harnessing the power of GATs, our method captures subtle cell relationships, offering highly accurate patient profiling. Based on evaluation of data from 30 patients, HemaGraph demonstrates classification performance across five different cell classes, outperforming traditional methodologies and state-of-the-art methods. Moreover, the uniqueness of this framework lies in the training and testing phase of HemaGraph, where it has been applied for extremely large graphs, containing up to hundreds of thousands of nodes and two million edges, to detect low frequency cell populations (e.g. 0.01% for one population), with accuracies reaching 98%. Our findings underscore the potential of HemaGraph in improving hematoligic multi-class classification, paving the way for patient-personalized interventions. To the best of our knowledge, this is the first effort to use GATs, and Graph Neural Networks (GNNs) in general, to classify cell populations from single-cell flow cytometry data. We envision applying this method to single-cell data from larger cohort of patients and on other hematologic diseases.

The time-consuming task of manual segmentation challenges routine systematic quantification of disease burden. Convolutional neural networks (CNNs) hold significant promise to reliably identify locations and boundaries of tumors from PET scans. We aimed to leverage the need for annotated data via semi-supervised approaches, with application to PET images of diffuse large B-cell lymphoma (DLBCL) and primary mediastinal large B-cell lymphoma (PMBCL). We analyzed 18F-FDG PET images of 292 patients with PMBCL (n=104) and DLBCL (n=188) (n=232 for training and validation, and n=60 for external testing). We employed FCM and MS losses for training a 3D U-Net with different levels of supervision: i) fully supervised methods with labeled FCM (LFCM) as well as Unified focal and Dice loss functions, ii) unsupervised methods with Robust FCM (RFCM) and Mumford-Shah (MS) loss functions, and iii) Semi-supervised methods based on FCM (RFCM+LFCM), as well as MS loss in combination with supervised Dice loss (MS+Dice). Unified loss function yielded higher Dice score (mean +/- standard deviation (SD)) (0.73 +/- 0.03; 95% CI, 0.67-0.8) compared to Dice loss (p-value<0.01). Semi-supervised (RFCM+alpha*LFCM) with alpha=0.3 showed the best performance, with a Dice score of 0.69 +/- 0.03 (95% CI, 0.45-0.77) outperforming (MS+alpha*Dice) for any supervision level (any alpha) (p<0.01). The best performer among (MS+alpha*Dice) semi-supervised approaches with alpha=0.2 showed a Dice score of 0.60 +/- 0.08 (95% CI, 0.44-0.76) compared to another supervision level in this semi-supervised approach (p<0.01). Semi-supervised learning via FCM loss (RFCM+alpha*LFCM) showed improved performance compared to supervised approaches. Considering the time-consuming nature of expert manual delineations and intra-observer variabilities, semi-supervised approaches have significant potential for automated segmentation workflows.

Deep learning models for medical image analysis often suffer from performance degradation when applied to data from different scanners or protocols, a phenomenon known as domain shift. This study investigates this challenge in the context of sex classification from 3D T1-weighted brain magnetic resonance imaging (MRI) scans using the IXI and OASIS3 datasets. While models achieved high within-domain accuracy (around 0.95) when trained and tested on a single dataset (IXI or OASIS3), we demonstrate a significant performance drop to chance level (about 0.50) when models trained on one dataset are tested on the other, highlighting the presence of a strong domain shift. To address this, we employed the ComBat harmonization technique to align the feature distributions of the two datasets. We evaluated three state-of-the-art 3D deep learning architectures (3D ResNet18, 3D DenseNet, and 3D EfficientNet) across multiple training strategies. Our results show that ComBat harmonization effectively reduces the domain shift, leading to a substantial improvement in cross-domain classification performance. For instance, the cross-domain balanced accuracy of our best model (ResNet18 3D with Attention) improved from approximately 0.50 (chance level) to 0.61 after harmonization. t-SNE visualization of extracted features provides clear qualitative evidence of the reduced domain discrepancy post-harmonization. This work underscores the critical importance of domain adaptation techniques for building robust and generalizable neuroimaging AI models.

Guidewire segmentation during endovascular interventions holds the potential to significantly enhance procedural accuracy, improving visualization and providing critical feedback that can support both physicians and robotic systems in navigating complex vascular pathways. Unlike supervised segmentation networks, which need many expensive expert-annotated labels, sim-to-real domain adaptation approaches utilize synthetic data from simulations, offering a cost-effective solution. The success of models like Segment-Anything (SAM) has driven advancements in image segmentation foundation models with strong zero/few-shot generalization through prompt engineering. However, they struggle with medical images like X-ray fluoroscopy and the domain-shifts of the data. Given the challenges of acquiring annotation and the accessibility of labeled simulation data, we propose a sim-to-real domain adaption framework with a coarse-to-fine strategy to adapt SAM to X-ray fluoroscopy guidewire segmentation without any annotation on the target domain. We first generate the pseudo-labels by utilizing a simple source image style transfer technique that preserves the guidewire structure. Then, we develop a weakly supervised self-training architecture to fine-tune an end-to-end student SAM with the coarse labels by imposing consistency regularization and supervision from the teacher SAM network. We validate the effectiveness of the proposed method on a publicly available Cardiac dataset and an in-house Neurovascular dataset, where our method surpasses both pre-trained SAM and many state-of-the-art domain adaptation techniques by a large margin. Our code will be made public on GitHub soon.

Brain tumor segmentation is highly contributive in diagnosing and treatment planning. The manual brain tumor delineation is a time-consuming and tedious task and varies depending on the radiologists skill. Automated brain tumor segmentation is of high importance, and does not depend on either inter or intra-observation. The objective of this study is to automate the delineation of brain tumors from the FLAIR, T1 weighted, T2 weighted, and T1 weighted contrast-enhanced MR sequences through a deep learning approach, with a focus on determining which MR sequence alone or which combination thereof would lead to the highest accuracy therein.

In the complex landscape of hematologic samples such as peripheral blood or bone marrow, cell classification, delineating diverse populations into a hierarchical structure, presents profound challenges. This study presents LeukoGraph, a recently developed framework designed explicitly for this purpose employing graph attention networks (GATs) to navigate hierarchical classification (HC) complexities. Notably, LeukoGraph stands as a pioneering effort, marking the application of graph neural networks (GNNs) for hierarchical inference on graphs, accommodating up to one million nodes and millions of edges, all derived from flow cytometry data. LeukoGraph intricately addresses a classification paradigm where for example four different cell populations undergo flat categorization, while a fifth diverges into two distinct child branches, exemplifying the nuanced hierarchical structure inherent in complex datasets. The technique is more general than this example. A hallmark achievement of LeukoGraph is its F-score of 98%, significantly outclassing prevailing state-of-the-art methodologies. Crucially, LeukoGraph's prowess extends beyond theoretical innovation, showcasing remarkable precision in predicting both flat and hierarchical cell types across flow cytometry datasets from 30 distinct patients. This precision is further underscored by LeukoGraph's ability to maintain a correct label ratio, despite the inherent challenges posed by hierarchical classifications.

Clinical SPECT-MPI images of 345 patients acquired from a dedicated cardiac SPECT in list-mode format were retrospectively employed to predict normal-dose images from low-dose data at the half, quarter, and one-eighth-dose levels. A generative adversarial network was implemented to predict non-gated normal-dose images in the projection space at the different reduced dose levels. Established metrics including the peak signal-to-noise ratio (PSNR), root mean squared error (RMSE), and structural similarity index metrics (SSIM) in addition to Pearson correlation coefficient analysis and derived parameters from Cedars-Sinai software were used to quantitatively assess the quality of the predicted normal-dose images. For clinical evaluation, the quality of the predicted normal-dose images was evaluated by a nuclear medicine specialist using a seven-point (-3 to +3) grading scheme. By considering PSNR, SSIM, and RMSE quantitative parameters among the different reduced dose levels, the highest PSNR (42.49) and SSIM (0.99), and the lowest RMSE (1.99) were obtained at the half-dose level in the reconstructed images. Pearson correlation coefficients were measured 0.997, 0.994, and 0.987 for the predicted normal-dose images at the half, quarter, and one-eighth-dose levels, respectively. Regarding the normal-dose images as the reference, the Bland-Altman plots sketched for the Cedars-Sinai selected parameters exhibited remarkably less bias and variance in the predicted normal-dose images compared with the low-dose data at the entire reduced dose levels. Overall, considering the clinical assessment performed by a nuclear medicine specialist, 100%, 80%, and 11% of the predicted normal-dose images were clinically acceptable at the half, quarter, and one-eighth-dose levels, respectively.

A standard dose of radioactive tracer must be delivered into the patients body to obtain high-quality Positron Emission Tomography (PET) images for diagnostic purposes, which raises the risk of radiation harm. A reduced tracer dose, on the other hand, results in poor image quality and a noise-induced quantitative bias in PET imaging. The partial volume effect (PVE), which is the result of PET intrinsic limited spatial resolution, is another source of quality and quantity degradation in PET imaging. The utilization of anatomical information for PVE correction (PVC) is not straightforward due to the internal organ motions, patient involuntary motions, and discrepancies in the appearance and size of the structures in anatomical and functional images. Furthermore, an additional MR imaging session is necessary for anatomical information, which may not be available. We set out to build a deep learning-based framework for predicting partial volume corrected full-dose (FD-PVC) pictures from either standard or low-dose (LD) PET images without requiring any anatomical data in order to provide a joint solution for PVC and denoise low-dose PET images.

We set out to simulate four reduced dose-levels (60%-dose, 40%-dose, 20%-dose, and 10%-dose) of standard CT imaging using Beer-Lambert's law across 49 patients infected with COVID-19. Then, three denoising filters, namely Gaussian, Bilateral, and Median, were applied to the different low-dose CT images, the quality of which was assessed prior to and after the application of the various filters via calculation of peak signal-to-noise ratio (PSNR), root mean square error (RMSE), structural similarity index measure (SSIM), and relative CT-value bias, separately for the lung tissue and whole-body. The quantitative evaluation indicated that 10%-dose CT images have inferior quality (with RMSE=322.1-+104.0 HU and bias=11.44-+4.49% in the lung) even after the application of the denoising filters. The bilateral filter exhibited superior performance to suppress the noise and recover the underlying signals in low-dose CT images compared to the other denoising techniques. The bilateral filter led to RMSE and bias of 100.21-+16.47 HU, -0.21-+1.20%, respectively in the lung regions for 20%-dose CT images compared to the Gaussian filter with RMSE=103.46-+15.70 HU and bias=1.02-+1.68%, median filter with RMSE=129.60-+18.09 HU and bias=-6.15-+2.24%, and the nonfiltered 20%-dose CT with RMSE=217.37-+64.66 HU and bias=4.30-+1.85%. In conclusion, the 20%-dose CT imaging followed by the bilateral filtering introduced a reasonable compromise between image quality and patient dose reduction.

The presence of metal implants within CT imaging causes severe attenuation of the X-ray beam. Due to the incomplete information recorded by CT detectors, artifacts in the form of streaks and dark bands would appear in the resulting CT images. The metal-induced artifacts would firstly affect the quantitative accuracy of CT imaging, and consequently, the radiation treatment planning and dose estimation in radiation therapy. To address this issue, CT scanner vendors have implemented metal artifact reduction (MAR) algorithms to avoid such artifacts and enhance the overall quality of CT images. The orthopedic-MAR (OMAR) and normalized MAR (NMAR) algorithms are the most well-known metal artifact reduction (MAR) algorithms, used worldwide. These algorithms have been implemented on Philips and Siemens scanners, respectively. In this study, we set out to quantitatively and qualitatively evaluate the effectiveness of these two MAR algorithms and their impact on accurate radiation treatment planning and CT-based dosimetry. The quantitative metrics measured on the simulated metal artifact dataset demonstrated superior performance of the OMAR technique over the NMAR one in metal artifact reduction. The analysis of radiation treatment planning using the OMAR and NMAR techniques in the corrected CT images showed that the OMAR technique reduced the toxicity of healthy tissues by 10% compared to the uncorrected CT images.

Objective: In this work, we set out to investigate the accuracy of direct attenuation correction (AC) in the image domain for the myocardial perfusion SPECT imaging (MPI-SPECT) using two residual (ResNet) and UNet deep convolutional neural networks. Methods: The MPI-SPECT 99mTc-sestamibi images of 99 participants were retrospectively examined. UNet and ResNet networks were trained using SPECT non-attenuation corrected images as input and CT-based attenuation corrected SPECT images (CT-AC) as reference. The Chang AC approach, considering a uniform attenuation coefficient within the body contour, was also implemented. Quantitative and clinical evaluation of the proposed methods were performed considering SPECT CT-AC images of 19 subjects as reference using the mean absolute error (MAE), structural similarity index (SSIM) metrics, as well as relevant clinical indices such as perfusion deficit (TPD). Results: Overall, the deep learning solution exhibited good agreement with the CT-based AC, noticeably outperforming the Chang method. The ResNet and UNet models resulted in the ME (count) of ${-6.99\pm16.72}$ and ${-4.41\pm11.8}$ and SSIM of ${0.99\pm0.04}$ and ${0.98\pm0.05}$, respectively. While the Change approach led to ME and SSIM of ${25.52\pm33.98}$ and ${0.93\pm0.09}$, respectively. Similarly, the clinical evaluation revealed a mean TPD of ${12.78\pm9.22}$ and ${12.57\pm8.93}$ for the ResNet and UNet models, respectively, compared to ${12.84\pm8.63}$ obtained from the reference SPECT CT-AC images. On the other hand, the Chang approach led to a mean TPD of ${16.68\pm11.24}$. Conclusion: We evaluated two deep convolutional neural networks to estimate SPECT-AC images directly from the non-attenuation corrected images. The deep learning solutions exhibited the promising potential to generate reliable attenuation corrected SPECT images without the use of transmission scanning.

Radiomics features extract quantitative information from medical images, towards the derivation of biomarkers for clinical tasks, such as diagnosis, prognosis, or treatment response assessment. Different image discretization parameters (e.g. bin number or size), convolutional filters, segmentation perturbation, or multi-modality fusion levels can be used to generate radiomics features and ultimately signatures. Commonly, only one set of parameters is used; resulting in only one value or flavour for a given RF. We propose tensor radiomics (TR) where tensors of features calculated with multiple combinations of parameters (i.e. flavours) are utilized to optimize the construction of radiomics signatures. We present examples of TR as applied to PET/CT, MRI, and CT imaging invoking machine learning or deep learning solutions, and reproducibility analyses: (1) TR via varying bin sizes on CT images of lung cancer and PET-CT images of head & neck cancer (HNC) for overall survival prediction. A hybrid deep neural network, referred to as TR-Net, along with two ML-based flavour fusion methods showed improved accuracy compared to regular rediomics features. (2) TR built from different segmentation perturbations and different bin sizes for classification of late-stage lung cancer response to first-line immunotherapy using CT images. TR improved predicted patient responses. (3) TR via multi-flavour generated radiomics features in MR imaging showed improved reproducibility when compared to many single-flavour features. (4) TR via multiple PET/CT fusions in HNC. Flavours were built from different fusions using methods, such as Laplacian pyramids and wavelet transforms. TR improved overall survival prediction. Our results suggest that the proposed TR paradigm has the potential to improve performance capabilities in different medical imaging tasks.

In the first step, a pre-trained model (YOLO) was used to detect all suspicious nod-ules. The YOLO model was re-trained using 397 CT images to detect the entire nodule in CT images. To maximize the sensitivity of the model, a confidence level (the probability threshold for object detection) of 0.3 was set for nodule detection in the first phase (ensuring the entire suspicious nodules are detected from the input CT images). The aim of the hierarchy model is to detect and classify the entire lung nodules (from CT images) with a low false-negative rate. Given the outcome of the first step, we proposed a 3D CNN classifier to analyze and classify the suspicious nodules detected by the YOLO model to achieve a nodule detection framework with a very low false-negative rate. This framework was evaluated using the LUNA 16 dataset, which consists of 888 CT images containing the location of 1186 nodules and 400000 non-nodules in the lung. Results: A large number of false positives were detected due to the low confidence level used in the YOLO model. Utilizing the 3D classifier, the accuracy of nodule detection was remarkably enhanced. The YOLO model detected 294 suspicious nodules (out of 321) when using a confidence level of 50%, wherein there were 107 false positives (187 true positives). By reducing the confidence level to 30%, 459 suspicious nodules were identified by the YOLO model, wherein 138 were false positives, and 321 were true positives. When the outcome of the YOLO model with a confidence level of 30% was fed into the 3D CNN classifier, a nodule detection accuracy of 98.4% and AUC of 98.9% were achieved Conclusion: The proposed framework resulted in a few false-negative and false-positives pre-dictions in nodule detection from CT images. The proposed approach would be helpful in detecting pulmonary nodules from CT images as a decision support tool.