Multiple sclerosis (MS) is common in adults while myelin oligodendrocyte glycoprotein antibody-associated disease (MOGAD) is rare. Our previous machine-learning algorithm, using clinical variables, ≤6 brain lesions, and no Dawson fingers, achieved 79% accuracy, 78% sensitivity, and 80% specificity in distinguishing MOGAD from MS but lacked validation. The aim of this study was to (1) evaluate the clinical/MRI algorithm for distinguishing MS from MOGAD, (2) develop a deep learning (DL) model, (3) assess the benefit of combining both, and (4) identify key differentiators using probability attention maps (PAMs). This multicenter, retrospective, cross-sectional MAGNIMS study included scans from 19 centers. Inclusion criteria were as follows: adults with non-acute MS and MOGAD, with high-quality T2-fluid-attenuated inversion recovery and T1-weighted scans. Brain scans were scored by 2 readers to assess the performance of the clinical/MRI algorithm on the validation data set. A DL-based classifier using a ResNet-10 convolutional neural network was developed and tested on an independent validation data set. PAMs were generated by averaging correctly classified attention maps from both groups, identifying key differentiating regions. We included 406 MRI scans (218 with relapsing remitting MS [RRMS], mean age: 39 years ±11, 69% F; 188 with MOGAD, age: 41 years ±14, 61% F), split into 2 data sets: a training/testing set (n = 265: 150 with RRMS, age: 39 years ±10, 72% F; 115 with MOGAD, age: 42 years ±13, 61% F) and an independent validation set (n = 141: 68 with RRMS, age: 40 years ±14, 65% F; 73 with MOGAD, age: 40 years ±15, 63% F). The clinical/MRI algorithm predicted RRMS over MOGAD with 75% accuracy (95% CI 67-82), 96% sensitivity (95% CI 88-99), and specificity 56% (95% CI 44-68) in the validation cohort. The DL model achieved 77% accuracy (95% CI 64-89), 73% sensitivity (95% CI 57-89), and 83% specificity (95% CI 65-96) in the training/testing cohort, and 70% accuracy (95% CI 63-77), 67% sensitivity (95% CI 55-79), and 73% specificity (95% CI 61-83) in the validation cohort without retraining. When combined, the classifiers reached 86% accuracy (95% CI 81-92), 84% sensitivity (95% CI 75-92), and 89% specificity (95% CI 81-96). PAMs identified key region volumes: corpus callosum (1872 mm³), left precentral gyrus (341 mm³), right thalamus (193 mm³), and right cingulate cortex (186 mm³) for identifying RRMS and brainstem (629 mm³), hippocampus (234 mm³), and parahippocampal gyrus (147 mm³) for identifying MOGAD. Both classifiers effectively distinguished RRMS from MOGAD. The clinical/MRI model showed higher sensitivity while the DL model offered higher specificity, suggesting complementary roles. Their combination improved diagnostic accuracy, and PAMs revealed distinct damage patterns. Future prospective studies should validate these models in diverse, real-world settings. This study provides Class III evidence that both a clinical/MRI algorithm and an MRI-based DL model accurately distinguish RRMS from MOGAD.

The potential of deep learning for medical imaging is often constrained by limited data availability. Generative models can unlock this potential by generating synthetic data that reproduces the statistical properties of real data while being more accessible for sharing. In this study, we investigated the influence of training set size on the performance of a state-of-the-art generative adversarial network, the StyleGAN2-ADA, trained on a cohort of 3,227 subjects from the OpenBHB dataset to generate 2D slices of brain MR images from healthy subjects. The quality of the synthetic images was assessed through qualitative evaluations and state-of-the-art quantitative metrics, which are provided in a publicly accessible repository. Our results demonstrate that StyleGAN2-ADA generates realistic and high-quality images, deceiving even expert radiologists while preserving privacy, as it did not memorize training images. Notably, increasing the training set size led to slight improvements in fidelity metrics. However, training set size had no noticeable impact on diversity metrics, highlighting the persistent limitation of mode collapse. Furthermore, we observed that diversity metrics, such as coverage and β-recall, are highly sensitive to the number of synthetic images used in their computation, leading to inflated values when synthetic data significantly outnumber real ones. These findings underscore the need to carefully interpret diversity metrics and the importance of employing complementary evaluation strategies for robust assessment. Overall, while StyleGAN2-ADA shows promise as a tool for generating privacy-preserving synthetic medical images, overcoming diversity limitations will require exploring alternative generative architectures or incorporating additional regularization techniques.

Pancreatic ductal adenocarcinoma (PDAC) is a leading cause of cancer-related deaths, with accurate staging being critical for treatment planning. Automated 3D segmentation models can aid in staging, but segmenting PDAC, especially in cases of locally advanced pancreatic cancer (LAPC), is challenging due to the tumor's heterogeneous appearance, irregular shapes, and extensive infiltration. This study developed and evaluated a tripartite self-supervised learning architecture for improved 3D segmentation of LAPC, addressing the challenges of heterogeneous appearance, irregular shapes, and extensive infiltration in PDAC. We implemented a tripartite architecture consisting of a teacher model, a professor model, and a student model. The teacher model, trained on manually segmented CT scans, generated initial pseudo-segmentations. The professor model refined these segmentations, which were then used to train the student model. We utilized 1115 CT scans from 903 patients for training. Three expert abdominal radiologists manually segmented 30 CT scans from 27 patients with LAPC, serving as reference standards. We evaluated the performance using DICE, Hausdorff distance (HD95), and mean surface distance (MSD). The teacher, professor, and student models achieved average DICE scores of 0.60, 0.73, and 0.75, respectively, with significant boundary accuracy improvements (teacher HD95/MSD, 25.71/5.96 mm; professor, 9.68/1.96 mm; student, 4.79/1.34 mm). Our findings demonstrate that the professor model significantly enhances segmentation accuracy for LAPC (p < 0.01). Both the professor and student models offer substantial improvements over previous work. The introduced tripartite self-supervised learning architecture shows promise for improving automated 3D segmentation of LAPC, potentially aiding in more accurate staging and treatment planning.

To explore the perspectives of AI vendors on the integration of AI in medical imaging and oncology clinical practice. An online survey was created on Qualtrics, comprising 23 closed and 5 open-ended questions. This was administered through social media, personalised emails, and the channels of the European Society of Medical Imaging Informatics and Health AI Register, to all those working at a company developing or selling accredited AI solutions for medical imaging and oncology. Quantitative data were analysed using SPSS software, version 28.0. Qualitative data were summarised using content analysis on NVivo, version 14. In total, 83 valid responses were received, with participants having a global distribution and diverse roles and professional backgrounds (business/management/clinical practitioners/engineers/IT, etc). The respondents mentioned the top enablers (practitioner acceptance, business case of AI applications, explainability) and challenges (new regulations, practitioner acceptance, business case) of AI implementation. Co-production with end-users was confirmed as a key practice by most (52.9%). The respondents recognised infrastructure issues within clinical settings (64.1%), lack of clinician engagement (54.7%), and lack of financial resources (42.2%) as key challenges in meeting customer expectations. They called for appropriate reimbursement, robust IT support, clinician acceptance, rigorous regulation, and adequate user training to ensure the successful integration of AI into clinical practice. This study highlights that people, infrastructure, and funding are fundamentals of AI implementation. AI vendors wish to work closely with regulators, patients, clinical practitioners, and other key stakeholders, to ensure a smooth transition of AI into daily practice. Question AI vendors' perspectives on unmet needs, challenges, and opportunities for AI adoption in medical imaging are largely underrepresented in recent research. Findings Provision of consistent funding, optimised infrastructure, and user acceptance were highlighted by vendors as key enablers of AI implementation. Clinical relevance Vendors' input and collaboration with clinical practitioners are necessary to clinically implement AI. This study highlights real-world challenges that AI vendors face and opportunities they value during AI implementation. Keeping the dialogue channels open is key to these collaborations.

<b>Objective</b>: Pharmacokinetic (PK) parameters derived from dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) provide quantitative characterization of tissue perfusion and permeability. However, existing deep learning methods for PK parameter estimation rely on either temporal or spatial features alone, overlooking the integrated spatial-temporal characteristics of DCE-MRI data. This study aims to remove this barrier by fully leveraging the spatial and temporal information to improve parameter estimation.&#xD;<b>Approach</b>: A spatial-temporal information-driven unsupervised deep learning method (STUDE) was proposed. STUDE combines convolutional neural networks (CNNs) and a customized Vision Transformer (ViT) to separately capture spatial and temporal features, enabling comprehensive modelling of contrast agent dynamics and tissue heterogeneity. Besides, a spatial-temporal attention (STA) feature fusion module was proposed to enable adaptive focus on both dimensions for more effective feature fusion. Moreover, the extended Tofts model imposed physical constraints on PK parameter estimation, enabling unsupervised training of STUDE. The accuracy and diagnostic value of STUDE was compared with the orthodox non-linear least squares (NLLS) and representative deep learning-based methods (i.e., GRU, CNN, U-Net, and VTDCE-Net) on a numerical brain phantom and 87 glioma patients, respectively.&#xD;<b>Main results</b>: On the numerical brain phantom, STUDE produced PK parameter maps with the lowest systematic and random errors even under low SNR conditions (SNR = 10 dB). On glioma data, STUDE generated parameter maps with reduced noise compared to NLLS and demonstrated superior structural clarity compared to other methods. Furthermore, STUDE outshined all other methods in the identification of glioma isocitrate dehydrogenase (IDH) mutation status, achieving the area under the curve (AUC) values at 0.840 and 0.908 for the receiver operating characteristic curves of<i>K^trans</i>and<i>V_e</i>, respectively. A combination of all PK parameters improved AUC to 0.926.&#xD;<b>Significance</b>: STUDE advances spatial-temporal information-driven and physics-informed learning for precise PK parameter estimation, demonstrating its potential clinical significance.&#xD.

Brain tumors represent a significant neurological challenge, affecting individuals across all age groups. Accurate and timely diagnosis of tumor types is critical for effective treatment planning. Magnetic Resonance Imaging (MRI) remains a primary diagnostic modality due to its non-invasive nature and ability to provide detailed brain imaging. However, traditional tumor classification relies on expert interpretation, which is time-consuming and prone to subjectivity. This study proposes a novel deep learning architecture, the Dual-Feature Cross-Fusion Network (DF-CFN), for the automated classification of brain tumors using MRI data. The model integrates ConvNeXt for capturing global contextual features and a shallow CNN combined with Feature Channel Attention Network (FcaNet) for extracting local features. These are fused through a cross-feature fusion mechanism for improved classification. The model is trained and validated using a Kaggle dataset encompassing four tumor classes (glioma, meningioma, pituitary, and non-tumor), achieving an accuracy of 99.33%. Its generalizability is further confirmed using the Figshare dataset, yielding 99.22% accuracy. Comparative analyses with baseline and recent models validate the superiority of DF-CFN in terms of precision and robustness. This approach demonstrates strong potential for assisting clinicians in reliable brain tumor classification, thereby improving diagnostic efficiency and reducing the burden on healthcare professionals.

This study compares various preprocessing techniques for hyperspectral deep learning-based cancer diagnostics. The study considers different spectrum scaling and noise reduction options across spatial and spectral axes of hyperspectral datacubes, as well varying levels of blood and light reflections removal. We also examine how the size of the patches extracted from the hyperspectral data affects the models' performance. We additionally explore various strategies to mitigate our dataset's imbalance (where cancerous tissues are underrepresented). Our results indicate that. Scaling: Standardization significantly improves both sensitivity and specificity compared to Normalization. Larger input patch sizes enhance performance by capturing more spatial context. Noise reduction unexpectedly degrades performance. Blood filtering is more effective than filtering reflected light pixels, although neither approach produces significant results. By carefully maintaining consistent testing conditions, we ensure a fair comparison across preprocessing methods and reproducibility. Our findings highlight the necessity of careful preprocessing selection to maximize deep learning performance in medical imaging applications.

People with semantic dementia (SD) or semantic variant primary progressive aphasia typically present with marked atrophy of the anterior temporal lobe, and thereafter progress more slowly than other forms of frontotemporal dementia. This suggests a prolonged prodromal phase with accumulation of neuropathology and minimal symptoms, about which little is known. To study early and presymptomatic SD, we first examine a well-characterised cohort of people with SD recruited from the Cambridge Centre for Frontotemporal Dementia. Five people with early SD had coincidental MRI prior to the onset of symptoms, or were healthy volunteers in research with anterior temporal lobe atrophy as an incidental finding. We model longitudinal imaging changes in left- and right-lateralised SD to predict atrophy at symptom onset. We then assess 61,203 participants with structural brain MRI in the UK Biobank to find individuals with imaging changes in keeping with SD but with no neurodegenerative diagnosis. To identify these individuals in UK Biobank, we design an ensemble-based classifier, differentiating baseline structural MRI in SD from healthy controls and patients with other neurodegenerative diseases, including other causes of frontotemporal lobar degeneration. We train the classifier on a Cambridge-based cohort (SD n=47, other neurodegenerative diseases n=498, healthy controls n=88) and test it on a combined cohort from the Neuroimaging in Frontotemporal Dementia study and the Alzheimer's Disease Neuroimaging Initiative (SD n=42, other neurodegenerative n=449, healthy control n=127). From our case series, we find people with marked atrophy three to five years before recognition of symptom onset in left- or right-predominant SD. We present right-lateralised cases with subtle multimodal semantic impairment, found concurrently with only mild behavioural disturbance. We show that imaging measures can be used to reliably and accurately differentiate clinical SD from other neurodegenerative diseases (recall 0.88, precision 0.95, F1 score 0.91). We find individuals with no neurodegenerative diagnosis in the UK Biobank with striking left-lateralised (prevalence ages 45-85 4.8/100,000) or right-lateralised (5.9/100,000) anterior temporal lobe atrophy, with deficits on cognitive testing suggestive of semantic impairment. These individuals show progressive involvement of other cognitive domains in longitudinal follow-up. Together, our findings suggest that (i) there is a burden of incipient early anterior temporal lobe atrophy in older populations, with comparable prevalence of left- and right-sided cases from this prospective unbiased approach to identification, (ii) substantial atrophy is required for manifest symptoms, particularly in right-lateralised cases, and (iii) semantic deficits across multiple domains can be detected in the early symptomatic phase.

Accurate ischemic stroke lesion segmentation is useful to define the optimal reperfusion treatment and unveil the stroke etiology. Despite the importance of diffusion-weighted MRI (DWI) for stroke diagnosis, learning from multi-sequence MRI images like apparent diffusion coefficient (ADC) can capitalize on the complementary nature of information from various modalities and show strong potential to improve the performance of segmentation. However, existing deep learning-based methods require large amounts of well-annotated data from multiple modalities for training, while acquiring such datasets is often impractical. We conduct the exploration of semi-supervised stroke lesion segmentation from multi-sequence MRI images by utilizing unlabeled data to improve performance using limited annotation and propose a novel framework by exploiting cross-modality collaboration and discrepancy to efficiently utilize unlabeled data. Specifically, we adopt a cross-modal bidirectional copy-paste strategy to enable information collaboration between different modalities and a cross-modal discrepancy-informed correction strategy to efficiently learn from limited labeled multi-sequence MRI data and abundant unlabeled data. Extensive experiments on the ischemic stroke lesion segmentation (ISLES 22) dataset demonstrate that our method efficiently utilizes unlabeled data with 12.32% DSC improvements compared with a supervised baseline using 10% annotations and outperforms existing semi-supervised segmentation methods with better performance.

Segmenting ischemic stroke lesions from Non-Contrast CT (NCCT) scans is a complex task due to the hypo-intense nature of these lesions compared to surrounding healthy brain tissue and their iso-intensity with lateral ventricles in many cases. Identifying early acute ischemic stroke lesions in NCCT remains particularly challenging. Computer-assisted detection and segmentation can serve as valuable tools to support clinicians in stroke diagnosis. This paper introduces CoAt U SegNet, a novel deep learning model designed to detect and segment acute ischemic stroke lesions from NCCT scans. Unlike conventional 3D segmentation models, this study presents an advanced 3D deep learning approach to enhance delineation accuracy. Traditional machine learning models have struggled to achieve satisfactory segmentation performance, highlighting the need for more sophisticated techniques. For model training, 50 NCCT scans were used, with 10 scans for validation and 500 scans for testing. The encoder convolution blocks incorporated dilation rates of 1, 3, and 5 to capture multi-scale features effectively. Performance evaluation on 500 unseen NCCT scans yielded a Dice similarity score of 75% and a Jaccard index of 70%, demonstrating notable improvement in segmentation accuracy. An enhanced similarity index was employed to refine lesion segmentation, which can further aid in distinguishing the penumbra from the core infarct area, contributing to improved clinical decision-making.