The Alzheimer′s Disease Prediction Of Longitudinal Evolution (TADPOLE) Challenge is the most comprehensive challenge to date with regard to the number of subjects, considered features, and challenge participants. The initial objective of TADPOLE was the identification of the most predictive data, features, and methods for the progression of subjects at risk of developing Alzheimer′s. The challenge was successful in recognizing tree-based ensemble methods such as gradient boosting and random forest as the best methods for the prognosis of the clinical status in Alzheimer’s disease (AD). However, the challenge outcome was limited to which combination of data processing and methods exhibits the best accuracy; hence, it is difficult to determine the contribution of the methods to the accuracy. The quantification of feature importance was globally approached by all the challenge participant methods. In addition, TADPOLE provided general answers that focused on improving performance while ignoring important issues such as interpretability. The purpose of this study is to intensively explore the models of the top three TADPOLE Challenge methods in a common framework for fair comparison. In addition, for these models, the most meaningful features for the prognosis of the clinical status of AD are studied and the contribution of each feature to the accuracy of the methods is quantified. We provide plausible explanations as to why the methods achieve such accuracy, and we investigate whether the methods use information coherent with clinical knowledge. Finally, we approach these issues through the analysis of SHapley Additive exPlanations (SHAP) values, a technique that has recently attracted increasing attention in the field of explainable artificial intelligence (XAI).

Background: Several studies indicate that the anterior visual pathway provides information about the dynamics of axonal degeneration in Multiple Sclerosis (MS). Current research in the field is focused on the quest for the most discriminative features among patients and controls and the development of machine learning models that yield computer-aided solutions widely usable in clinical practice. However, most studies are conducted with small samples and the models are used as black boxes. Clinicians should not trust machine learning decisions unless they come with comprehensive and easily understandable explanations. Materials and methods: A total of 216 eyes from 111 healthy controls and 100 eyes from 59 patients with relapsing-remitting MS were enrolled. The feature set was obtained from the thickness of the ganglion cell layer (GCL) and the retinal nerve fiber layer (RNFL). Measurements were acquired by the novel Posterior Pole protocol from Spectralis Optical Coherence Tomography (OCT) device. We compared two black-box methods (gradient boosting and random forests) with a glass-box method (explainable boosting machine). Explainability was studied using SHAP for the black-box methods and the scores of the glass-box method. Results: The best-performing models were obtained for the GCL layer. Explainability pointed out to the temporal location of the GCL layer that is usually broken or thinning in MS and the relationship between low thickness values and high probability of MS, which is coherent with clinical knowledge. Conclusions: The insights on how to use explainability shown in this work represent a first important step toward a trustworthy computer-aided solution for the diagnosis of MS with OCT.

Computational Anatomy aims for the study of the statistical variability in anatomical structures. Variability is encoded by the transformations existing among populations of anatomical images. These transformations are usually computed from diffeomorphic registration based on the large deformation paradigm. In this framework diffeomorphisms are usually computed as end points of paths on the Riemannian manifold of diffeomorphisms parameterized by non-stationary vector fields. Recently, an alternative parameterization based on stationary vector fields has been developed. In this article we propose to use this stationary parameterization for diffeomorphic registration. We formulate the variational problem related to this registration scenario and derive the associated Euler-Lagrange equations. We evaluate the performance of the non-stationary vs the stationary parameterizations in real and synthetic 3D-MRI datasets. Compared to the non-stationary parameterization, our proposal provides similar accuracy in terms of image matching and deformation smoothness while drastically reducing memory and time requirements.

Deep learning has achieved expert-level performance in medical imaging analysis. However, models often fail to generalize across patient populations due to cross-population domain shifts, distributional differences arising from demographic variability, variations in imaging protocols, scanner hardware, and differences in disease prevalence. This challenge limits the real-world deployment and can increase health inequities. This review systematically examines the nature, causes, and impact of cross-population domain shift in deep learning-based medical imaging analysis. We analyzed 50 peer-reviewed studies from 2020 to 2025, evaluating the proposed methodologies for handling population shifts, the datasets employed, and the metrics used to assess performance. Our findings demonstrate that performance degradation ranged from 10–25% when models were tested on unseen populations, emphasizing the substantial impact of domain shifts on model generalizability. The literature reveals that mitigation strategies broadly fall into two categories: data-centric approaches, such as augmentation and harmonization, and model-centric approaches, including domain adaptation, transfer learning, adversarial learning, multi-task learning, and continual learning. While domain adaptation and transfer learning are the most widely used, their performance gains across populations remain modest, ranging from 5–15%, and are not supported by external validation. Our synthesis reveals a significant reliance on large, publicly available datasets from limited regions, with an underrepresentation of data from low- and middle-income countries. Evaluation practices are inconsistent, with few studies employing standardized external test sets. This review provides a structured taxonomy of mitigation techniques, a refined analysis of domain shift characteristics, and an in-depth critique of methodological challenges. We highlight the urgent need for more geographically and demographically inclusive datasets, adaptable modeling techniques, and standardized evaluation protocols to enable accurate and equitable AI-driven diagnostics across diverse populations. Finally, we outline future research directions to guide the development of robust, generalizable, and fair models for medical imaging analysis.