Foundation models in medical imaging have shown promising label efficiency, achieving high performance on downstream tasks using only a fraction of the annotated data otherwise required. In this study, we evaluate this potential in the context of prostate multiparametric MRI using ProFound, a recently developed domain-specific vision foundation model pretrained on large-scale prostate MRI datasets. We investigate the impact of variable image quality on the label-efficient finetuning, by quantifying the generalisability of the finetuned models. We conduct a comprehensive set of experiments by systematically varying the ratios of high- and low-quality images in the finetuning and evaluation sets. Our findings indicate that image quality distribution and its finetune-and-test mismatch significantly affect model performance. In particular: a) Varying the ratio of high- to low-quality images between finetuning and test sets leads to notable differences in downstream performance; and b) The presence of sufficient high-quality images in the finetuning set is critical for maintaining strong performance, whilst the importance of matched finetuning and testing distribution varies between different downstream tasks, such as automated radiology reporting and prostate cancer detection. Importantly, experimental results also show that, although finetuning requires significantly less labeled data compared to training from scratch when the quality ratio is consistent, this label efficiency is not independent of the image quality distribution. For example, we show cases that, without sufficient high-quality images in finetuning, finetuned models may fail to outperform those without pretraining.

For adult diffuse gliomas (ADGs), most grading can be achieved through molecular subtyping, retaining only two key histopathological features for high-grade glioma (HGG): necrosis (NEC) and microvascular proliferation (MVP). We developed a deep learning (DL) framework to automatically identify and characterize these features. We trained patch-level models to detect and quantify NEC and MVP using a dataset that employed active learning, incorporating patches from 621 whole-slide images (WSIs) from the Chinese Glioma Genome Atlas (CGGA). Utilizing trained patch-level models, we effectively integrated the predicted outcomes and positions of individual patches within WSIs from The Cancer Genome Atlas (TCGA) cohort to form datasets. Subsequently, we introduced a patient-level model, named PLNet (Probability Localization Network), which was trained on these datasets to facilitate patient diagnosis. We also explored the subtypes of NEC and MVP based on the features extracted from patch-level models with clustering process applied on all positive patches. The patient-level models demonstrated exceptional performance, achieving an AUC of 0.9968, 0.9995 and AUPRC of 0.9788, 0.9860 for NEC and MVP, respectively. Compared to pathological reports, our patient-level models achieved the accuracy of 88.05% for NEC and 90.20% for MVP, along with a sensitivity of 73.68% and 77%. When sensitivity was set at 80%, the accuracy for NEC reached 79.28% and for MVP reached 77.55%. DL models enabled more efficient and accurate histopathological image analysis which will aid traditional glioma diagnosis. Clustering-based analyses utilizing features extracted from patch-level models could further investigate the subtypes of NEC and MVP.

Modern deep neural networks are highly over-parameterized, necessitating the use of data augmentation techniques to prevent overfitting and enhance generalization. Generative adversarial networks (GANs) are popular for synthesizing visually realistic images. However, these synthetic images often lack diversity and may have ambiguous class labels. Recent data mixing strategies address some of these issues by mixing image labels based on salient regions. Since the main diagnostic information is not always contained within the salient regions, we aim to address the resulting label mismatches in medical image classifications. We propose a variety-guided data mixing framework (VariMix), which exploits an absolute difference map (ADM) to address the label mismatch problems of mixed medical images. VariMix generates ADM using the image-to-image (I2I) GAN across multiple classes and allows for bidirectional mixing operations between the training samples. The proposed VariMix achieves the highest accuracy of 99.30% and 94.60% with a SwinT V2 classifier on a Chest X-ray (CXR) dataset and a Retinal dataset, respectively. It also achieves the highest accuracy of 87.73%, 99.28%, 95.13%, and 95.81% with a ConvNeXt classifier on a Breast Ultrasound (US) dataset, a CXR dataset, a Retinal dataset, and a Maternal-Fetal US dataset, respectively. Furthermore, the medical expert evaluation on generated images shows the great potential of our proposed I2I GAN in improving the accuracy of medical image classifications. Extensive experiments demonstrate the superiority of VariMix compared with the existing GAN- and Mixup-based methods on four public datasets using Swin Transformer V2 and ConvNeXt architectures. Furthermore, by projecting the source image to the hyperplanes of the classifiers, the proposed I2I GAN can generate hyperplane difference maps between the source image and the hyperplane image, demonstrating its ability to interpret medical image classifications. The source code is provided in https://github.com/yXiangXiong/VariMix.

Multiple sclerosis (MS) is a chronic inflammatory neurodegenerative disorder of the central nervous system (CNS) and represents the leading cause of non-traumatic disability among young adults. Magnetic resonance imaging (MRI) has revolutionized both the clinical management and scientific understanding of MS, serving as an indispensable paraclinical tool. Its high sensitivity and diagnostic accuracy enable early detection and timely therapeutic intervention, significantly impacting patient outcomes. Recent technological advancements have facilitated the integration of artificial intelligence (AI) algorithms for automated lesion identification, segmentation, and longitudinal monitoring. The ongoing refinement of deep learning (DL) and machine learning (ML) techniques, alongside their incorporation into clinical workflows, holds great promise for improving healthcare accessibility and quality in MS management. Despite the encouraging performance of DL models in MS lesion segmentation and disease progression tracking, their effectiveness is frequently constrained by the scarcity of large, diverse, and publicly available datasets. Open-source initiatives such as MSLesSeg, MS-Baghdad, MS-Shift, and MSSEG-2 have provided valuable contributions to the research community. Building upon these foundations, we introduce the SibBMS dataset to further advance data-driven research in MS. In this study, we present the SibBMS dataset, a carefully curated, open-source resource designed to support MS research utilizing structural brain MRI. The dataset comprises imaging data from 93 patients diagnosed with MS or radiologically isolated syndrome (RIS), alongside 100 healthy controls. All lesion annotations were manually delineated and rigorously reviewed by a three-tier panel of experienced neuroradiologists to ensure clinical relevance and segmentation accuracy. Additionally, the dataset includes comprehensive demographic metadata--such as age, sex, and disease duration--enabling robust stratified analyses and facilitating the development of more generalizable predictive models. Our dataset is available via a request-access form at https://forms.gle/VqTenJ4n8S8qvtxQA.

Artificial intelligence (AI) holds significant potential to enhance various aspects of oncology, spanning the cancer care continuum. This review provides an overview of current and emerging AI applications, from risk assessment and early detection to treatment and supportive care. AI-driven tools are being developed to integrate diverse data sources, including multi-omics and electronic health records, to improve cancer risk stratification and personalize prevention strategies. In screening and diagnosis, AI algorithms show promise in augmenting the accuracy and efficiency of medical image analysis and histopathology interpretation. AI also offers opportunities to refine treatment planning, optimize radiation therapy, and personalize systemic therapy selection. Furthermore, AI is explored for its potential to improve survivorship care by tailoring interventions and to enhance end-of-life care through improved symptom management and prognostic modeling. Beyond care delivery, AI augments clinical workflows, streamlines the dissemination of up-to-date evidence, and captures critical patient-reported outcomes for clinical decision support and outcomes assessment. However, the successful integration of AI into clinical practice requires addressing key challenges, including rigorous validation of algorithms, ensuring data privacy and security, and mitigating potential biases. Effective implementation necessitates interdisciplinary collaboration and comprehensive education for health care professionals. The synergistic interaction between AI and clinical expertise is crucial for realizing the potential of AI to contribute to personalized and effective cancer care. This review highlights the current state of AI in oncology and underscores the importance of responsible development and implementation.

Surgeons treating skeletally immature patients use skeletal age to determine appropriate surgical strategies. Traditional bone age estimation methods utilizing hand radiographs are time-consuming. To develop highly accurate/reliable deep learning (DL) models for determination of accurate skeletal age from hand radiographs. Cohort Study. The authors utilized 3 publicly available hand radiograph data sets for model development/validation from (1) the Radiological Society of North America (RSNA), (2) the Radiological Hand Pose Estimation (RHPE) data set, and (3) the Digital Hand Atlas (DHA). All 3 data sets report corresponding sex and skeletal age. The RHPE and DHA also contain chronological age. After image preprocessing, a ConvNeXt model was trained first on the RSNA data set using sex/skeletal age as inputs using 5-fold cross-validation, with subsequent training on the RHPE with addition of chronological age. Final model validation was performed on the DHA and an institutional data set of 200 images. The first model, trained on the RSNA, achieved a mean absolute error (MAE) of 3.68 months on the RSNA test set and 5.66 months on the DHA. This outperformed the 4.2 months achieved on the RSNA test set by the best model from previous work (12.4% improvement) and 3.9 months by the open-source software Deeplasia (5.6% improvement). After incorporation of chronological age from the RHPE in model 2, this error improved to an MAE of 4.65 months on the DHA, again surpassing the best previously published models (19.8% improvement). Leveraging newer DL technologies trained on >20,000 hand radiographs across 3 distinct, diverse data sets, this study developed a robust model for predicting bone age. Utilizing features extracted from an RSNA model, combined with chronological age inputs, this model outperforms previous state-of-the-art models when applied to validation data sets. These results indicate that the models provide a highly accurate/reliable platform for clinical use to improve confidence about appropriate surgical selection (eg, physeal-sparing procedures) and time savings for orthopaedic surgeons/radiologists evaluating skeletal age. Development of an accurate DL model for determination of bone age from the hand reduces the time required for age estimation. Additionally, streamlined skeletal age estimation can aid practitioners in determining optimal treatment strategies and may be useful in research settings to decrease workload and improve reporting.

Ultrasound imaging is critical for clinical diagnostics, providing insights into various diseases and organs. However, artificial intelligence (AI) in this field faces challenges, such as the need for large labeled datasets and limited task-specific model applicability, particularly due to ultrasound's low signal-to-noise ratio (SNR). To overcome these, we introduce the Ultrasound Representation Foundation Model (URFM), designed to learn robust, generalizable representations from unlabeled ultrasound images, enabling label-efficient adaptation to diverse diagnostic tasks. URFM is pre-trained on over 1M images from 15 major anatomical organs using representation-based masked image modeling (MIM), an advanced self-supervised learning. Unlike traditional pixel-based MIM, URFM integrates high-level representations from BiomedCLIP, a specialized medical vision-language model, to address the low SNR issue. Extensive evaluation shows that URFM outperforms state-of-the-art methods, offering enhanced generalization, label efficiency, and training-time efficiency. URFM's scalability and flexibility signal a significant advancement in diagnostic accuracy and clinical workflow optimization in ultrasound imaging.

Non-contrast CT (NCCT) is widely used in clinical practice and holds potential for large-scale atherosclerosis screening, yet its application in detecting and grading aortic atherosclerosis remains limited. To address this, we propose Aortic-AAE, an automated segmentation system based on a cascaded attention mechanism within the nnU-Net framework. The cascaded attention module enhances feature learning across complex anatomical structures, outperforming existing attention modules. Integrated preprocessing and post-processing ensure anatomical consistency and robustness across multi-center data. Trained on 435 labeled NCCT scans from three centers and validated on 388 independent cases, Aortic-AAE achieved 81.12% accuracy in aortic stenosis classification and 92.37% in Agatston scoring of calcified plaques, surpassing five state-of-the-art models. This study demonstrates the feasibility of using deep learning for accurate detection and grading of aortic atherosclerosis from NCCT, supporting improved diagnostic decisions and enhanced clinical workflows.

The present study examined an integrated multiple neuroimaging modality (T1 structural, Diffusion Tensor Imaging (DTI), and resting-state FMRI (rsFMRI)) to predict aphasia severity using Western Aphasia Battery-Revised Aphasia Quotient (WAB-R AQ) in 76 individuals with post-stroke aphasia. We employed Support Vector Regression (SVR) and Random Forest (RF) models with supervised feature selection and a stacked feature prediction approach. The SVR model outperformed RF, achieving an average root mean square error (RMSE) of 16.38±5.57, Pearson's correlation coefficient (r) of 0.70±0.13, and mean absolute error (MAE) of 12.67±3.27, compared to RF's RMSE of 18.41±4.34, r of 0.66±0.15, and MAE of 14.64±3.04. Resting-state neural activity and structural integrity emerged as crucial predictors of aphasia severity, appearing in the top 20% of predictor combinations for both SVR and RF. Finally, the feature selection method revealed that functional connectivity in both hemispheres and between homologous language areas is critical for predicting language outcomes in patients with aphasia. The statistically significant difference in performance between the model using only single modality and the optimal multi-modal SVR/RF model (which included both resting-state connectivity and structural information) underscores that aphasia severity is influenced by factors beyond lesion location and volume. These findings suggest that integrating multiple neuroimaging modalities enhances the prediction of language outcomes in aphasia beyond lesion characteristics alone, offering insights that could inform personalized rehabilitation strategies.

Cortical morphological abnormalities in schizophrenia (SCZ), major depressive disorder (MDD), and bipolar disorder (BD) have been identified in past research. However, their potential as objective biomarkers to differentiate these disorders remains uncertain. Machine learning models may offer a novel diagnostic tool. Structural MRI (sMRI) of 220 SCZ, 220 MDD, 220 BD, and 220 healthy controls were obtained using a 3T scanner. Volume, thickness, surface area, and mean curvature of 68 cerebral cortices were extracted using FreeSurfer. 272 features underwent 3 feature selection techniques to isolate important variables for model construction. These features were incorporated into 3 classifiers for classification. After model evaluation and hyperparameter tuning, the best-performing model was identified, along with the most significant brain measures. The univariate feature selection-Naive Bayes model achieved the best performance, with an accuracy of 0.66, macro-average AUC of 0.86, and sensitivities and specificities ranging from 0.58-0.86 to 0.81-0.93, respectively. Key features included thickness of right isthmus-cingulate cortex, area of left inferior temporal gyrus, thickness of right superior temporal gyrus, mean curvature of right pars orbitalis, thickness of left transverse temporal cortex, volume of left caudal anterior-cingulate cortex, area of right banks superior temporal sulcus, and thickness of right temporal pole. The machine learning model based on sMRI data shows promise for aiding in the differential diagnosis of SCZ, MDD, and BD. Cortical features from the cingulate and temporal lobes may highlight distinct biological mechanisms underlying each disorder.