Deep learning-based medical image classification techniques are rapidly advancing in medical image analysis, making it crucial to develop accurate and trustworthy models that can be efficiently deployed across diverse clinical scenarios. Concept Bottleneck Models (CBMs), which first predict a set of explainable concepts from images and then perform classification based on these concepts, are increasingly being adopted for explainable medical image classification. However, the inherent explainability of CBMs introduces new challenges when deploying trained models to new environments. Variations in imaging protocols and staining methods may induce concept-level shifts, such as alterations in color distribution and scale. Furthermore, since CBM training requires explicit concept annotations, fine-tuning models solely with image-level labels could compromise concept prediction accuracy and faithfulness - a critical limitation given the high cost of acquiring expert-annotated concept labels in medical domains. To address these challenges, we propose a training-free confusion concept identification strategy. By leveraging minimal new data (e.g., 4 images per class) with only image-level labels, our approach enhances out-of-domain performance without sacrificing source domain accuracy through two key operations: masking misactivated confounding concepts and amplifying under-activated discriminative concepts. The efficacy of our method is validated on both skin and white blood cell images. Our code is available at: https://github.com/riverback/TF-TTI-XMed.

Aimed to develop and validate a CT radiomics-based explainable machine learning model for diagnosing malignancy and benignity specifically in endometrial cancer (EC) patients. A total of 83 EC patients from two centers, including 46 with malignant and 37 with benign conditions, were included, with data split into a training set (n=59) and a testing set (n=24). The regions of interest (ROIs) were manually segmented from pre-surgical CT scans, and 1132 radiomic features were extracted from the pre-surgical CT scans using Pyradiomics. Six explainable machine learning modeling algorithms were implemented respectively, for determining the optimal radiomics pipeline. The diagnostic performance of the radiomic model was evaluated by using sensitivity, specificity, accuracy, precision, F1 score, confusion matrices, and ROC curves. To enhance clinical understanding and usability, we separately implemented SHAP analysis and feature mapping visualization, and evaluated the calibration curve and decision curve. By comparing six modeling strategies, the Random Forest model emerged as the optimal choice for diagnosing EC, with a training AUC of 1.00 and a testing AUC of 0.96. SHAP identified the most important radiomic features, revealing that all selected features were significantly associated with EC (P < 0.05). Radiomics feature maps also provide a feasible assessment tool for clinical applications. DCA indicated a higher net benefit for our model compared to the "All" and "None" strategies, suggesting its clinical utility in identifying high-risk cases and reducing unnecessary interventions. In conclusion, the CT radiomics-based explainable machine learning model achieved high diagnostic performance, which could be used as an intelligent auxiliary tool for the diagnosis of endometrial cancer.

Thyroid cancer is among the most common cancers in the United States. Thyroid nodules are frequently detected through ultrasound (US) imaging, and some require further evaluation via fine-needle aspiration (FNA) biopsy. Despite its effectiveness, FNA often leads to unnecessary biopsies of benign nodules, causing patient discomfort and anxiety. To address this, the American College of Radiology Thyroid Imaging Reporting and Data System (TI-RADS) has been developed to reduce benign biopsies. However, such systems are limited by interobserver variability. Recent deep learning approaches have sought to improve risk stratification, but they often fail to utilize the rich temporal and spatial context provided by US cine clips, which contain dynamic global information and surrounding structural changes across various views. In this work, we propose the Spatio-Temporal Cross Attention for Cine Thyroid Ultrasound Time Series Classification (STACT-Time) model, a novel representation learning framework that integrates imaging features from US cine clips with features from segmentation masks automatically generated by a pretrained model. By leveraging self-attention and cross-attention mechanisms, our model captures the rich temporal and spatial context of US cine clips while enhancing feature representation through segmentation-guided learning. Our model improves malignancy prediction compared to state-of-the-art models, achieving a cross-validation precision of 0.91 (plus or minus 0.02) and an F1 score of 0.89 (plus or minus 0.02). By reducing unnecessary biopsies of benign nodules while maintaining high sensitivity for malignancy detection, our model has the potential to enhance clinical decision-making and improve patient outcomes.

With the advancement of Large Language Model (LLM) for natural language processing, this paper presents an intriguing finding: a frozen pre-trained LLM layer can process visual tokens for medical image segmentation tasks. Specifically, we propose a simple hybrid structure that integrates a pre-trained, frozen LLM layer within the CNN encoder-decoder segmentation framework (LLM4Seg). Surprisingly, this design improves segmentation performance with a minimal increase in trainable parameters across various modalities, including ultrasound, dermoscopy, polypscopy, and CT scans. Our in-depth analysis reveals the potential of transferring LLM's semantic awareness to enhance segmentation tasks, offering both improved global understanding and better local modeling capabilities. The improvement proves robust across different LLMs, validated using LLaMA and DeepSeek.

Synthetic data has recently reached a level of visual fidelity that makes it nearly indistinguishable from real data, offering great promise for privacy-preserving data sharing in medical imaging. However, fully synthetic datasets still suffer from significant limitations: First and foremost, the legal aspect of sharing synthetic data is often neglected and data regulations, such as the GDPR, are largley ignored. Secondly, synthetic models fall short of matching the performance of real data, even for in-domain downstream applications. Recent methods for image generation have focused on maximising image diversity instead of fidelity solely to improve the mode coverage and therefore the downstream performance of synthetic data. In this work, we shift perspective and highlight how maximizing diversity can also be interpreted as protecting natural persons from being singled out, which leads to predicate singling-out (PSO) secure synthetic datasets. Specifically, we propose a generalisable framework for training diffusion models on personal data which leads to unpersonal synthetic datasets achieving performance within one percentage point of real-data models while significantly outperforming state-of-the-art methods that do not ensure privacy. Our code is available at https://github.com/MischaD/Trichotomy.

Federated learning has significantly advanced distributed training of machine learning models across decentralized data sources. However, existing frameworks often lack comprehensive solutions that combine uncertainty quantification, interpretability, and robustness. To address this, we propose FedNAM+, a federated learning framework that integrates Neural Additive Models (NAMs) with a novel conformal prediction method to enable interpretable and reliable uncertainty estimation. Our method introduces a dynamic level adjustment technique that utilizes gradient-based sensitivity maps to identify key input features influencing predictions. This facilitates both interpretability and pixel-wise uncertainty estimates. Unlike traditional interpretability methods such as LIME and SHAP, which do not provide confidence intervals, FedNAM+ offers visual insights into prediction reliability. We validate our approach through experiments on CT scan, MNIST, and CIFAR datasets, demonstrating high prediction accuracy with minimal loss (e.g., only 0.1% on MNIST), along with transparent uncertainty measures. Visual analysis highlights variable uncertainty intervals, revealing low-confidence regions where model performance can be improved with additional data. Compared to Monte Carlo Dropout, FedNAM+ delivers efficient and global uncertainty estimates with reduced computational overhead, making it particularly suitable for federated learning scenarios. Overall, FedNAM+ provides a robust, interpretable, and computationally efficient framework that enhances trust and transparency in decentralized predictive modeling.

This study was undertaken to develop and validate an automatic, artificial intelligence-enhanced software tool for hippocampal sclerosis (HS) detection, using a variety of standard magnetic resonance imaging (MRI) protocols from different MRI scanners for routine clinical practice. First, MRI scans of 36 epilepsy patients with unilateral HS and 36 control patients with epilepsy of other etiologies were analyzed. MRI features, including hippocampal subfield volumes from three-dimensional (3D) magnetization-prepared rapid acquisition gradient echo (MPRAGE) scans and fluid-attenuated inversion recovery (FLAIR) intensities, were calculated. Hippocampal subfield volumes were corrected for total brain volume and z-scored using a dataset of 256 healthy controls. Hippocampal subfield FLAIR intensities were z-scored in relation to each subject's mean cortical FLAIR signal. Additionally, left-right ratios of FLAIR intensities and volume features were obtained. Support vector classifiers were trained on the above features to predict HS presence and laterality. In a second step, the algorithm was validated using two independent, external cohorts, including 118 patients and 116 controls in sum, scanned with different MRI scanners and acquisition protocols. Classifiers demonstrated high accuracy in HS detection and lateralization, with slight variations depending on the input image availability. The best cross-validation accuracy was achieved using both 3D MPRAGE and 3D FLAIR scans (mean accuracy = 1.0, confidence interval [CI] = .939-1.0). External validation of trained classifiers in two independent cohorts yielded accuracies of .951 (CI = .902-.980) and .889 (CI = .805-.945), respectively. In both validation cohorts, the additional use of FLAIR scans led to significantly better classification performance than the use of MPRAGE data alone (p = .016 and p = .031, respectively). A further model was trained on both validation cohorts and tested on the former training cohort, providing additional evidence for good validation performance. Comparison to a previously published algorithm showed no significant difference in performance (p = 1). The method presented achieves accurate automated HS detection using standard clinical MRI protocols. It is robust and flexible and requires no image processing expertise.

Diseases underlying osteolytic lesions in jaws are characterized by the absorption of bone tissue and are often asymptomatic, delaying their diagnosis. Well-defined lesions (benign cyst-like lesions) and ill-defined lesions (osteomyelitis or malignancy) can be detected early in a panoramic radiograph (PR) by an experienced examiner, but most dentists lack appropriate training. To support dentists, this study aimed to develop and evaluate deep learning models for the detection of osteolytic lesions in PRs. A dataset of 676 PRs (165 well-defined, 181 ill-defined, 330 control) was collected from the Department of Oral and Maxillofacial Surgery at Charité Berlin, Germany. The osteolytic lesions were pixel-wise segmented and labeled as well-defined or ill-defined. Four model architectures for instance segmentation (Mask R-CNN with a Swin-Tiny or ResNet-50 backbone, Mask DINO, and YOLOv5) were employed with five-fold cross-validation. Their effectiveness was evaluated with sensitivity, specificity, F1-score, and AUC and failure cases were shown. Mask R-CNN with a Swin-Tiny backbone was most effective (well-defined F1 = 0.784, AUC = 0.881; ill-defined F1 = 0.904, AUC = 0.971) and the model architectures including vision transformer components were more effective than those without. Model mistakes were observed around the maxillary sinus, at tooth extraction sites, and for radiolucent bands. Promising deep learning models were developed for the detection of osteolytic lesions in PRs, particularly those with vision transformer components (Mask R-CNN with Swin-Tiny and Mask DINO). These results underline the potential of vision transformers for enhancing the automated detection of osteolytic lesions, offering a significant improvement over traditional deep learning models.

The therapeutic management of neoplasm employs high level energy beam to ablate malignant cells, which can cause collateral damage to adjacent normal tissue. Furthermore, respiration-induced organ motion, during radiotherapy can lead to significant displacement of neoplasms. In this work, a non-invasive ultrasound-based deep learning algorithm for respiratory motion compensation system (RMCS) was developed to mitigate the effect of respiratory motion induced neoplasm movement in radiotherapy. The deep learning algorithm generated based on modified YOLOv8n (You Only Look Once), by incorporating squeeze and excitation blocks for channel wise recalibration and enhanced attention mechanisms for spatial channel focus (SE-ATT-YOLO) to cope up with enhanced ultrasound image detection in real time scenario. The trained model was inferred with ultrasound movement of human diaphragm and tracked the bounding box coordinates using BoT-Sort, which drives the RMCS. The SE-ATT-YOLO model achieved mean average precision (mAP) of 0.88 which outperforms YOLOv8n with the value of 0.85. The root mean square error (RMSE) obtained from prerecorded respiratory signals with the compensated RMCS signal was calculated. The model achieved an inference speed of approximately 50 FPS. The RMSE values recorded were 4.342 for baseline shift, 3.105 for sinusoidal signal, 1.778 for deep breath, and 1.667 for slow signal. The SE-ATT-YOLO model outperformed all the results of previous models. The loss function uncertainty in YOLOv8n model was rectified in SE-ATT YOLO depicting the stability of the model. The model' stability, speed and accuracy of the model optimized the performance of the RMCS.

To develop a radiomics risk-predictive model for differentiating the different stages of Hashimoto's thyroiditis (HT). Data from patients with HT who underwent definitive surgical pathology between January 2018 and December 2023 were retrospectively collected and categorized into early HT (HT patients with simple positive antibodies or simultaneously accompanied by elevated thyroid hormones) and late HT (HT patients with positive antibodies and beginning to present subclinical hypothyroidism or developing hypothyroidism). Ultrasound images and five clinical and 12 laboratory indicators were obtained. Six classifiers were used to construct radiomics models. The gradient boosting decision tree (GBDT) classifier was used to screen for the best features to explore the main risk factors for differentiating early HT. The performance of each model was evaluated by receiver operating characteristic (ROC) curve. The model was validated using one internal and two external test cohorts. A total of 785 patients were enrolled. Extreme gradient boosting (XGBOOST) showed best performance in the training cohort, with an AUC of 0.999 (0.998, 1), and AUC values of 0.993 (0.98, 1), 0.947 (0.866, 1), and 0.98 (0.939, 1), respectively, in the internal test, first external, and second external cohorts. Ultrasound radiomic features contributed to 78.6% (11/14) of the model. The first-order feature of traverse section of thyroid ultrasound image, texture feature gray-level run length matrix (GLRLM) of longitudinal section of thyroid ultrasound image and free thyroxine showed the greatest contributions in the model. Our study developed and tested a risk-predictive model that effectively differentiated HT stages to more precisely and actively manage patients with HT at an earlier stage.