Counterfactual explanations (CEs) aim to enhance the interpretability of machine learning models by illustrating how alterations in input features would affect the resulting predictions. Common CE approaches require an additional model and are typically constrained to binary counterfactuals. In contrast, we propose a novel method that operates directly on the latent space of a generative model, specifically a Diffusion Autoencoder (DAE). This approach offers inherent interpretability by enabling the generation of CEs and the continuous visualization of the model’s internal representation across decision boundaries. Our method leverages the DAE’s ability to encode images into a semantically rich latent space in an unsupervised manner, eliminating the need for labeled data or separate feature extraction models. We show that these latent representations are helpful for medical condition classification and the ordinal regression of severity pathologies, such as vertebral compression fractures (VCF) and diabetic retinopathy (DR). Beyond binary CEs, our method supports the visualization of ordinal CEs using a linear model, providing deeper insights into the model’s decision-making process and enhancing interpretability. Experiments across various medical imaging datasets demonstrate the method’s advantages in interpretability and versatility. The linear manifold of the DAE’s latent space allows for meaningful interpolation and manipulation, making it a powerful tool for exploring medical image properties. Our code is available at https://github.com/matanat/dae_counterfactual

Accurate calibration of finite element (FE) models is essential across various biomechanical applications, including human intervertebral discs (IVDs), to ensure their reliability and use in diagnosing and planning treatments. However, traditional calibration methods are computationally intensive, requiring iterative, derivative-free optimization algorithms that often take days to converge. This study addresses these challenges by introducing a novel, efficient, and effective calibration method demonstrated on a human L4-L5 IVD FE model as a case study using a neural network (NN) surrogate. The NN surrogate predicts simulation outcomes with high accuracy, outperforming other machine learning models, and significantly reduces the computational cost associated with traditional FE simulations. Next, a Projected Gradient Descent (PGD) approach guided by gradients of the NN surrogate is proposed to efficiently calibrate FE models. Our method explicitly enforces feasibility with a projection step, thus maintaining material bounds throughout the optimization process. The proposed method is evaluated against state-of-the-art Genetic Algorithm (GA) and inverse model baselines on synthetic and in vitro experimental datasets. Our approach demonstrates superior performance on synthetic data, achieving a Mean Absolute Error (MAE) of 0.06 compared to the baselines' MAE of 0.18 and 0.54, respectively. On experimental specimens, our method outperforms the baseline in 5 out of 6 cases. While our approach requires initial dataset generation and surrogate training, these steps are performed only once, and the actual calibration takes under three seconds. In contrast, traditional calibration time scales linearly with the number of specimens, taking up to 8 days in the worst-case. Such efficiency paves the way for applying more complex FE models, potentially extending beyond IVDs, and enabling accurate patient-specific simulations.

Chronic rhinosinusitis (CRS) is a common and persistent sinus inflammation that affects 5%-12% of the general population. It substantially reduces quality of life, yet its severity is often challenging to assess objectively. The Lund-Mackay score (LMS) rates sinus opacification but is typically assessed manually and subjectively. We introduce Paranasal Segmentation for Imaging-based Disease Evaluation (PARASIDE), an automatic tool for segmenting air and soft tissue volumes of the structures of the sinus maxillaris, frontalis, sphenoidalis, and ethmoidalis in T1-weighted magnetic resonance imaging. Utilizing that segmentation, we quantify feature relations such as volume, thickness, and intensity relations which were previously observed only manually and subjectively. Using these features, we regress the Total Lund-Mackay Score (TLMS) of each subject. We compare our approach against established baselines: the Quantitative Opacification Score (QOS) and the Quantitative Lund-Mackay Score (QLMS). PARASIDE achieves a mean-squared error (MSE) of 2.444 and mean absolute error (MAE) of 1.181 for TLMS prediction, outperforming the QOS/QLMS baseline (MSE = 3.784, MAE = 1.445). The segmentation achieves a mean Dice similarity coefficient of 0.882 ± 0.138 and an average symmetric surface distance (ASSD) of 0.311 ± 0.354 mm across all structures. PARASIDE enables the first automated whole-paranasal sinus segmentation for T1-weighted MRI, extracting quantitative features that predict CRS severity more accurately than existing volumetric scoring methods. By integrating high-quality segmentation with fully automated TLMS estimation, our system offers a reproducible and objective assessment tool in clinical workflows, with the potential to reduce inter-rater variability, accelerate reporting, and support large-scale retrospective studies.

Thoracolumbar stump ribs are one of the essential indicators of thoracolumbar transitional vertebrae or enumeration anomalies. While some studies manually assess these anomalies and describe the ribs qualitatively, this study aims to automate thoracolumbar stump rib detection and analyze their morphology quantitatively. To this end, we train a high-resolution deep learning model for rib segmentation using nnUNet and achieve significant improvements over existing models (Dice score 0.997 vs. 0.779, p-value < 0.01). In addition, we employ a novel iterative algorithm and piecewise linear interpolation to estimate rib length, achieving a success rate of 98.2%. When analyzing morphological features, we show that stump ribs articulate more posteriorly at the vertebrae ($-19.2\pm 3.8$ vs. $-13.8\pm 2.5$ mm, p-value < 0.01), are thinner ($260.6\pm 103.4$ vs. $563.6\pm 127.1$${\mathrm{mm}}^2$, p-value < 0.01), and are oriented more downwards and sideways within the first centimeters in contrast to full-length ribs. We show that with partially visible ribs, these features can achieve an F1-score of 0.84 and an AUC of 0.98 in differentiating stump ribs from regular ones. We publish the model weights and masks for public use.

Individual intervertebral disc (IVD) characteristics contribute to spinal mechanics and pain, but individualized biomechanical modeling is insufficiently examined, especially regarding the validity of using a single material model across different degeneration grades and disc geometries. Based on MRI images, we generated and simulated 241 patient-specific lumbar finite element method (FEM) IVD models using a single material model calibrated to a healthy L4-L5 disc. This study is twofold: In one respect, it establishes and evaluates the automated morphing algorithm femReg, which is used to generate the models. In another respect, it evaluates the capacity of the geometry-only individualized IVD models to represent the range of motion (ROM) of variously degenerated discs by simulating them in four load cases under 1, 2.5, 5 and 7.5 Nm load. To do so, we grouped numerical models by height loss and compared absolute and normalized ROM results with in vitro data across different degeneration grades. 241 high-quality hexahedral IVD models (mean aspect ratio: 1.65, mean Hausdorff distance: 0.037 mm) were created and simulated within an average runtime of 6 minutes per IVD. Healthy models (n = 149) reproduced experimental ROM with root mean square errors (RMSEs) of 0.96^∘-2.05^∘ across bending load cases. The closest agreement was observed in lateral bending (RMSE = 0.68^∘), where facet joint contributions are minimal. In contrast, degenerated models exhibited reduced height-normalized ROM compared with in vitro data (RMSE = 0.05-0.24^∘/mm), indicating an effectively stiffer response and suggesting that degeneration-specific material parameters, such as fiber loosening or reduced hydration, should be incorporated to better capture pathological behavior.