Magnetic resonance imaging (MRI)-guided radiotherapy (MRIgRT) integrates MRI with linear accelerators (MRI-linacs), enabling real-time motion management based on temporally resolved 2D MRI (cine-MRI). Current systems rely on template matching or deformable image registration for radiotherapy target (typically the gross tumor volume) localization, which allows beam gating. Further advances in localization could support more precise and efficient delivery methods. https://trackrad2025.grand-challenge.org/ was organized to provide a common dataset to benchmark algorithms for MRIgRT target tracking in 2D+t cine-MRI. Participants propagated target segmentation masks from an initialization frame across subsequent frames. The dataset comprised sagittal cine-MRI scans of 585 cancer patients undergoing radiotherapy at 0.35 T and 1.5 T MRI-linacs at six different institutions, with expert-annotated targets in 108 sequences. Target sites included the thorax (179 cases), abdomen (266 cases), and pelvis (140 cases). A total of 477 unlabeled and 50 labeled cases were provided for training purposes, 58 cases were kept private for preliminary testing (8) and final evaluation (50). The algorithms submitted by participants were executed on the challenge platform and assessed using metrics in three categories: geometric accuracy, surrogate dose accuracy and execution speed. Rankings were derived via a Rank-Then-Mean scheme. TrackRAD2025 attracted 148 registrations from 28 countries, 100 preliminary submissions and 24 final submissions from 14 teams. The top five methods achieved mean Dice similarity coefficients >0.87 and Euclidean center distances <2.1 mm, comparable to interobserver variability. Leading top five solutions featured foundation models with (4) or without (1) finetuning. Field strength had minimal effect on performance and tracking worked better for the pelvis with reduced motion amplitude compared to the thorax and abdomen cases, which achieved equivalent performance. TrackRAD2025 established a benchmark for MRIgRT tracking on multi-institutional cine-MRI data, highlighting foundation models as promising for clinical translation.

Accurate delineation of pancreatic tumors on Magnetic Resonance Imaging (MRI) is important for diagnosis, radiotherapy treatment planning, and outcome assessment, but remains challenging due to complex anatomy and subtle tumor appearance. In routine practice, tumor contours on MRI are produced manually, which is time-consuming and subject to inter-observer variability. Radiotherapy on MRI-Linear Accelerator (MRI-Linac) systems further requires fast and consistent Gross Tumor Volume (GTV) contours for online adaptation, yet most public pancreas tumor segmentation benchmarks focus on Computed Tomography (CT). The Pancreatic Tumor Segmentation in Therapeutic and Diagnostic MRI (PANTHER) challenge addresses this gap by benchmarking automatic pancreatic tumor segmentation on MRI. The dataset includes contrast-enhanced T1-weighted diagnostic MRI and T2-weighted MRI-Linac scans with expert pancreas and tumor annotations, organized into two tasks: (1) tumor segmentation on diagnostic MRI and (2) tumor segmentation on MRI-Linac images. Performance was evaluated using overlap metrics, distance-based metrics, and tumor volume error. The challenge attracted 285 registered participants, with 12 and 9 final submissions for Tasks 1 and 2, respectively. On diagnostic MRI, top methods achieved performance close to inter-reader agreement. Multi-reader analysis suggested that models often reproduced the contouring style of the training annotator, highlighting the importance of annotation quality and consensus. In contrast, performance on MRI-Linac images was lower and more heterogeneous, including cases of complete localization failure. PANTHER provides the first public benchmark for pancreatic tumor segmentation on MRI, showing that clinically useful automation is feasible on diagnostic MRI, while robust MRI-Linac GTV segmentation remains an open challenge.

Over the past five years, artificial intelligence (AI) has introduced new models and methods for addressing the challenges associated with the broader adoption of AI models and systems in medicine. This paper reviews recent advances in AI for medical image and video analysis, outlines emerging paradigms, highlights pathways for successful clinical translation, and provides recommendations for future work. Hybrid Convolutional Neural Network (CNN) Transformer architectures now deliver state-of-the-art results in segmentation, classification, reconstruction, synthesis, and registration. Foundation and generative AI models enable the use of transfer learning to smaller datasets with limited ground truth. Federated learning supports privacy-preserving collaboration across institutions. Explainable and trustworthy AI approaches have become essential to foster clinician trust, ensure regulatory compliance, and facilitate ethical deployment. Together, these developments pave the way for integrating AI into radiology, pathology, and wider healthcare workflows.

Colorectal cancer (CRC) is a largely preventable disease influenced by modifiable behavioral risk factors such as diet, smoking, alcohol consumption, physical inactivity, and chronic stress. This study proposes a machine learning-based framework that generates personalized lifestyle recommendations aimed at CRC prevention. The system consists of two components: the Behavioral Recommendation Mapping Engine, which maps behavioral questionnaire responses to expertvalidated recommendations, and the Risk Assessment Module, which classifies participants into specific recommendations using supervised learning models. Eight domain-specific classifiers were developed, each targeting a key behavioral risk factor. Random Forests consistently outperformed Decision Trees, achieving high macro-averaged F1 scores even in imbalanced categories such as smoking$(F 1=0.88)$and stress$(F 1=0.80)$. The system also identifies the most influential behavioral variable per domain to highlight actionable risk factors. This framework will be integrated into the DIOPTRA mobile application to support realtime, personalized prevention.

E/MEG source imaging (ESI) provides non-invasive measurements of brain activity with high spatial and temporal resolution. In particular, the wearability and portability of EEG make it an attractive area of research beyond the biomedical communities, especially given the broad application prospects including brain-computer interface (BCI), neuromarketing and neuroergonomics. Although existing reviews offer valuable insights, they often present ESI models in a relatively isolated manner and may not encompass the most recent advancements in the field. In this work, we aim to: 1) provide a timely in-depth review of the widely-explored and state-of-the-art ESI models, including their underlying neurophysiological assumptions and mathematical derivations; 2) list the primary applications of ESI and highlight crucial steps regarding its implementations; 3) discuss current challenges in ESI and propose future research prospects; 4) demonstrate practical usage and implementation details of various representative ESI models. As a rapidly expanding field, ESI is continuously developing and evolving to integrate new technologies. We believe the widespread applications of ESI is happening, and it will dramatically expand our understanding of brain dynamics.

Breast cancer stands as the most prevalent cancer in women globally, with its worldwide escalating incidence and mortality rates underscoring the necessity of improving upon current non-invasive diagnostic methodologies for early-stage detection. This study introduces Rad-EfficientNet, a convolutional neural network (CNN) that incorporates radiomic features in its training pipeline to differentiate benign from malignant breast tumors in multiparametric 3 T breast magnetic resonance imaging (MRI). To this end, a dataset of 104 cases, including 45 benign and 59 malignant instances, was collected, and radiomic features were extracted from the 3D bounding boxes of each of the tumors. The Pearson's correlation coefficient and the Variance Inflation Factor were employed to reduce the radiomic features to a subset of 25. Rad-EfficientNet was then trained on both image and radiomics data. Based on the EfficientNet network family, the proposed Rad-EfficientNet architecture builds upon it by introducing a radiomics fusion layer consisting of a feature reduction operation, radiomic feature concatenation with the learned features, and finally a dropout layer. Rad-EfficientNet achieved an accuracy score of 82%, outperforming conventional classifiers trained solely on radiomic features, as well as hybrid models that combine learned and radiomic features post-training. These results indicate that by incorporating radiomics directly into the CNN training pipeline, complementary features are learned, thereby offering a way to improve current diagnostic deep learning techniques for breast lesion diagnosis.

The human brain connectome has long been recognized as a crucial component for various cognitive functions. While connectome-based predictive modeling (CPM) has been extensively explored for predicting behavior outcomes at the individual-level, its application to electroencephalogram (EEG) remains limited due to the inherent diversity and complexity of EEG frequency information. In the present work, we aim to address this issue by developing a filter bank CPM (FBCPM) framework that leverages narrowband EEG functional connectivity (FC) for individual prediction. Four independent datasets comprising 280 healthy subjects with 392 EEG recordings during the psychomotor vigilance test (PVT), were adopted here. Using the discovery dataset (i.e., Dataset 1) with 137 recordings, the feasibility of FBCPM was evaluated via predicting mean reaction time (RT) measures within a 15-min PVT task. The results showed that FBCPM framework achieved notable prediction accuracy and outperformed four benchmark approaches. Subsequent comprehensive internal and external validation analyses further affirmed its robustness across various hyper-parameters and generalizability to another three independent datasets (i.e., Dataset 2 to Dataset 4) with divergent recording or preprocessing settings. Moreover, the FBCPM framework exhibited satisfactory performance when generalized to time-on-task (TOT) effect measures (i.e., $\mathit {\Delta RT}$ and $\mathit {TOT_{slope}}$). Further investigation of contributing features to mean RT prediction indicated the remarkable predictive ability of negative features, manifesting as a pattern of low-frequency (below 8 Hz) predominance and complex topological distributions. Overall, these findings indicated that FBCPM provided a significant methodological advance in EEG-based individual prediction approaches, moving a step forward towards practical application in cognitive neuroscience.

Cardiac rehabilitation (CR) programs are vital for people recovering from cardiac surgeries or events. However, the effectiveness of CR programs varies and some patients may not adhere to them, which might result in less favourable outcomes. Machine Learning (ML) models could help predict the effectiveness and adherence of CR programs. This study proposes two such models: a) the CR program effectiveness prediction model and b) the CR program adherence prediction model. The models were trained on data from retrospective cohort study with 1448 participants collected at the Cardiac Rehabilitation Unit of the Hospital Clinico de Santiago de Compostela in Galicia, Spain (SERGAS). Data cleaning, normalization, imputation, statistical analysis, feature selection and repeated stratified k-fold cross-validation (CV) were applied on the ML pipeline, which tested and evaluated on baseline demographic, clinical, exercise tests and behavioral features. The performance of ML models was assessed by mean Area Under operating characteristic Curve (AUC), specificity, sensitivity, and balanced accuracy with 95% confidence interval (CI). The results show that Random Forest (RF) outperformed other evaluated classifiers for the CR program effectiveness model, with the highest AUC value of 0.789 (0.775, 0.802), while the best classifier for the CR adherence model was the Logistic Regression (LR) classifier, with an AUC value of 0.757 (0.749, 0.764). SHAP plots were also used to investigate the relationships among the variables used in the analysis. Finally, a two-dimensional scoring system was developed to jointly assess predicted adherence and effectiveness, enabling personalized visualization of patient response to CR.

Representation learning methods that discover task and/or data-specific characteristics are very popular for a variety of applications. However, their application to 3D medical images is restricted by the computational cost and their inherent subtle differences in intensities and appearance. In this paper, a novel representation learning scheme for extracting representations capable of distinguishing high-uptake regions from 3D 18F-Fluorodeoxyglucose positron emission tomography (FDG-PET) images is proposed. In particular, we propose a novel position-enhanced learning scheme effectively incorporating semantic and position-based features through our proposed Position Encoding Block (PEB) to produce highly informative representations. Such representations incorporate both semantic and position-aware features from high-dimensional medical data, leading to general representations with better performance on clinical tasks. To evaluate our method, we conducted experiments on the challenging task of classifying high-uptake regions as either non-tumor or tumor lesions in Metastatic Melanoma (MM). MM is a type of cancer characterized by its rapid spread to various body sites, which leads to low survival rates. Extensive experiments on an in-house and a public dataset of whole-body FDG-PET images indicated an increase of 10.50% in sensitivity and 4.89% in F1-score against the baseline representation learning scheme while also outperforming state-of-the-art methods for classifying MM regions of interest. The source code will be available at https://github.com/theoVag/Representation-Learning-Sem-Pos.

Heart failure (HF) is a complex syndrome that is affected by many factors and causes. It is crucial to early recognize the disease subtypes and the unidentified clinical pathways that give rise to it. Machine learning (ML) is the tool that assist to deal with these challenges and improve the prediction of HF. In this work, the HF risk prediction was implemented by employed ML classifiers, such as Random Forest (RF), Extreme Grading Boosting (XGBoost) and Light Gradient-Boosting Machine (LGBM). We utilized the data from the German epidemiological trial on ankle brachial index - getABI cohort, which includes 6,454 patients. The performance of classifiers was estimated by Accuracy (ACC), Sensitivity, Specificity and the area under the receiver operating characteristic curve (AUC) in mean values for each ML classifier. The results were also interpreted using the Explainable artificial intelligence (XAI) approach, the Shapley Additive exPlanations (SHAP) values. Our work reveals that LGBM classifier predict the HF risk within 5 years follow-up in general population with 68 % accuracy. Moreover, the N-terminal pro-B-type natriuretic peptide (NT-proBNP) was identified as the most important feature for HF risk prediction.

The identification of the location of prostate cancer is of paramount importance for improved treatment. This process is strictly bonded with the accurate segmentation of the prostate gland and its zones, on MR images. In the present study, an ensemble of 3 deep learning models along with a Meta-learner module able to refine the outcomes of the models, is proposed (Transi-Net) to segment the prostate's transition zone. A method to quantify the model's uncertainty is introduced to measure the confidence of an architecture with respect to its final decision. The backbone of Transi-Net consist the original U-net, Dense2U-net and Bridged U-net models. The proposed model showcased significant improvement in comparison with its base components as well as an independent model, the USE-Net, while it was proven more confident about its decision. The proposed model resulted in an improvement of 5%, 3%, 3% and 4% for Sensitivity, Balanced Accuracy, Dice Score and Rand Error Index respectively, compared to the second best, USE-Net.

As a common complaint in contemporary society, mental fatigue is a key element in the deterioration of the daily activities known as time-on-task (TOT) effect, making the prediction of fatigue-related performance decline exceedingly important. However, conventional group-level brain-behavioral correlation analysis has the limitation of generalizability to unseen individuals and fatigue prediction at individual-level is challenging due to the significant differences between individuals both in task performance efficiency and brain activities. Here, we introduced a cross-validated data-driven analysis framework to explore, for the first time, the feasibility of utilizing pre-task idiosyncratic resting-state functional connectivity (FC) on the prediction of fatigue-related task performance degradation at individual level. Specifically, two behavioral metrics, namely$\Delta$RT (between the most vigilant and fatigued states) and$TOT_{slope}$over the course of the 15-min sustained attention task, were estimated among three sessions from 37 healthy subjects to represent fatigue-related individual behavioral impairment. Then, a connectome-based prediction model was employed on pre-task resting-state FC features, identifying the network-related differences that contributed to the prediction of performance deterioration. As expected, prominent populational TOT-related performance declines were revealed across three sessions accompanied with substantial inter-individual differences. More importantly, we achieved significantly high accuracies for individualized prediction of both TOT-related behavioral impairment metrics using pre-task neuroimaging features. Despite the distinct patterns between both behavioral metrics, the identified top FC features contributing to the individualized predictions were mainly resided within/between frontal, temporal and parietal areas. Overall, our results of individualized prediction framework extended conventional correlation/classification analysis and may represent a promising avenue for the development of applicable techniques that allow precaution of the TOT-related performance declines in real-world scenarios.

Metastatic Melanoma (MM) is an aggressive type of cancer which produces metastases throughout the body with very poor survival rates. Recent advances in immunotherapy have shown promising results for controlling disease's progression. Due to the often rapid progression, fast and accurate diagnosis and treatment response assessment is vital for the whole patient management. These procedures prerequisite accurate, whole-body tumor identification. This can be offered by the imaging modality Positron Emission Tomography (PET)/Computed Tomography (CT) with the radiotracer F 18-Fluorodeoxyglucose (FDG). However, manual segmentation of PET/CT images is a very time-consuming and labor intensive procedure that requires expert knowledge. Most of the previously published segmentation techniques focus on a specific type of tumor or part of the body and require a great amount of manually labeled data, which is, however, difficult for MM. Multimodal analysis of PET/CT is also crucial because FDG-PET contains only the functional information of tumors which can be complemented by the anatomical information of CT. In this paper, we propose a whole-body segmentation framework capable of efficiently identifying the highly heterogeneous tumor lesions of MM from the whole-body 3D FDG-PET/CT images. The proposed decision support system begins with an Ensemble Unsupervised Segmentation of regions of high FDG-uptake based on Fuzzy C-means and a custom region growing algorithm. Then, a region classification model based on radiomics features and Neural Networks classifies these regions as tumors or not. Experimental results showed high performance in the identification of MM lesions with Sensitivity 83.68%, Specificity 91.82%, F1-score 75.42%, AUC 94.16% and Balanced accuracy 87.75% which were also supported by the public dataset evaluation.

The prediction of schizophrenia-related psychopathologic deficits is exceedingly important in the fields of psychiatry and clinical practice. However, objective association of the brain structure alterations to the illness clinical symptoms is challenging. Although, schizophrenia has been characterized as a brain dysconnectivity syndrome, evidence accounting for neuroanatomical network alterations remain scarce. Moreover, the absence of generalized connectome biomarkers for the assessment of illness progression further perplexes the prediction of long-term symptom severity. In this paper, a combination of individualized prediction models with quantitative graph theoretical analysis was adopted, providing a comprehensive appreciation of the extent to which the brain network properties are affected over time in schizophrenia. Specifically, Connectome-based Prediction Models were employed on Structural Connectivity (SC) features, efficiently capturing individual network-related differences, while identifying the anatomical connectivity disturbances contributing to the prediction of psychopathological deficits. Our results demonstrated distinctions among widespread cortical circuits responsible for different domains of symptoms, indicating the complex neural mechanisms underlying schizophrenia. Furthermore, the generated models were able to significantly predict changes of symptoms using SC features at follow-up, while the preserved SC features suggested an association with improved positive and overall symptoms. Moreover, cross-sectional significant deficits were observed in network efficiency and a progressive aberration of global integration in patients compared to healthy controls, representing a group-consensus pathological map, while supporting the dysconnectivity hypothesis.

In the nascent field of neuroergonomics, mental workload assessment is one of the most important issues and has an apparent significance in real-world applications. Although prior research has achieved efficient single-task classification, scatted studies on cross-task mental workload assessment usually result in unsatisfactory performance. Here, we introduce a data-driven analysis framework to overcome the challenges regarding task-independent workload assessment using a fusion of EEG spectral characteristics and unveil the common neural mechanisms underlying mental workload. Specifically, multi-frequency power spectrum and functional connectivity (FC) were estimated for two workload levels in two working-memory tasks performed by 40 healthy participants, subsequently being fed into a machine learning approach to obtain the importance of each feature vector and evaluate classification performance in a cross-task fashion. Our framework achieved a classification accuracy of 0.94 for task-independent mental workload discrimination. Further investigation of the designated features in terms of their spectral and localization properties revealed task-independent common patterns in the neural mechanisms governing workload. In particular, increased workload was associated with elevated frontal delta and theta power but reduced parietal alpha power, whereas FC exhibited complex frequency- and region-dependent alterations. By implication, the employment of the EEG feature fusion emphasized their utility in serving as promising indicators for different workload conditions applications.

Diagnosis of Autism Spectrum Disorder (ASD) is a complex task that typically relies on the expertise of the clinician due to the lack of specific quantitative biomarkers. As a consequence, automatic categorization of an individual within the ASD taxonomy poses many challenges, usually with controversial results. The implementation of Machine Learning approaches as a diagnostic tool for ASD classification is rapidly growing in the field of neuroscience, holding the potential to enhance discrimination validity among ASD and Typically Developed (TD) individuals, while providing indications in regard to ASD differentiating factors. In this study, various feature selection and classification techniques were employed in order to successfully discern between ASD and TD, using data from large resting-state functional Magnetic Resonance Imaging (rs-fMRI) database. Moreover, we adopt novel features, namely the Haralick texture features and the Kullback-Leibler divergence, combined with already established ones (i.e. static Functional Connectivity and demographics), assessing the most informative global attributes. Our framework succeeded in the identification of a small number of discriminative features, leading to high performance relative to previous works with optimal classification accuracy of 0.725.

MODELHealth is an innovative software platform aiming at developing an end-to-end solution for the process of pumping, enriching and anonymizing heterogeneous health data, and implementing Machine Learning (ML) methods for healthcare and research purposes that also leverage indoor localization services. The ultimate goal of the platform is to provide any authorized Information System with powerful algorithms and tools through Application Program Interfaces in order to facilitate and enhance clinical decision-making techniques at a broad range of healthcare services.

Trabecular bone fractures constitute a major health issue for the modern societies, with the currently established prediction methods of fracture risk, such as bone mineral density (BMD), resulting in errors up to 40%. Fracture-zone prediction based on bone's microstructure has been recently proposed as an alternative prediction method of fracture risk. In this paper, a classification system (CS) for the automatic fracture-zone prediction based on an Ensemble of Imbalanced Learning methods is proposed, following the observation that the percentage of the actual fractured bone area is significantly smaller than the intact bone in the case of a fracture event. The sample is divided into Volumes of Interest (VOIs) of specific size and 29 morphometrical parameters are calculated from each VOI, which serve as input features for the CS in order for it to separate the input patterns in to two classes: fractured and nonfractured. To this end, two well-established Imbalanced Learning methods, namely Random Undersampling and Synthetic Minority Oversampling, and two popular classification algorithms, namely Multilayer Perceptrons and Support Vector Machines, are tested and combined accordingly, to provide the best possible performance on a dataset that contains 45 specimens' pre- and postfailure scans. The best combination is then compared with three well-established Ensembles of Imbalanced Learning methods, namely RUSBoost, UnderBagging and SMOTEBagging. The experimental results clearly show that the proposed CS outperforms the competition, scoring in some occasions more than 90% in G-Mean and Area under Curve metrics. Finally, an investigation on the significance of the various trabecular bone's biomechanical parameters is made using the sequential forward floating selection technique, in order to identify possible biomarkers for fracture-zone prediction.

Prophylactic cranial irradiation (PCI) is a standard treatment technique for small cell lung cancer patients. However, there is evidence that this technique may contribute to neurocognitive deficits. Therefore the study of anatomical and functional connectivity in patients undergoing PCI as well as their neurocognitive functionality, depending on the type of disease and the phase of treatment and time of clinical examination, is of considerable interest. In this context, we investigate whether there are any differentiations in brain function during resting state and task-related functional magnetic resonance imaging (fMRI) in patients with cancer before PCI compared to healthy subjects. During a finger tapping task, the brain regions that were activated bilaterally for both groups are consistent with previous studies. During rest, the Default Mode Network (DMN) was identified in both groups. The preliminary results presented herein are subject to further investigation with larger patient and control group.

The diagnosis of thyroid malignancy by fine needle aspiration (FNA) examination has been proven to show wide variations of sensitivity and specificity. This paper proposes the utilization of a computer-aided diagnosis system based on a supervised classification algorithm from the artificial immune systems to assist the task of thyroid malignancy diagnosis. The core of the proposed algorithm is the so-called BoxCells, which are defined as parallelepipeds in the feature space. Properly defined operators act on the BoxCells in order to convert them into individual, elementary classifiers. The proposed algorithm is applied on FNA data from 2016 subjects with verified diagnosis and has exhibited average specificity higher than 99%, 90% sensitivity, and 98.5% accuracy. Furthermore, 24% of the cases that are characterized as "suspicious" by FNA and are histologically proven nonmalignancies have been classified correctly.

Event-Related Potentials (ERPs) provide non-invasive measurements of the electrical activity on the scalp related to the processing of stimuli and preparation of responses by the brain. In this paper, an ERP-signal classification method capable of discriminating between ERPs of correct and incorrect responses of actors is proposed. A number of histogram-related features were calculated from each ERP-signal and the most significant ones were extracted using the Sequential Forward Floating Selection algorithm along with the Fuzzy C-Means clustering algorithm. The Fuzzy C-Means algorithm was also used for the classification task. The approach yielded classification accuracy 93.75% for the actors’ correct and incorrect responses. The proposed ERP-signal classification method provides a promising tool to study error detection and observational-learning mechanisms in joint-action research and may foster the future development of systems capable of automatically detecting erroneous actions in human-human and human-artificial agent interactions.

In this paper, a digital subtraction radiology scheme is presented based on a new method for the automatic registration of dental radiographs acquired with or without rigorous a priori standardization. The scheme is comprised of an automatic registration method and a subtraction process. The proposed registration method can be considered as an object-based registration method without imposing the prerequisite of image segmentation in order to detect the boundary of the objects of interest or the automatic detection of matching landmarks. This is achieved by augmenting the dimensionality of the problem from two-dimensional gray-level matching to three-dimensional surface matching using the process of lifting in combination with a surface-matching technique. The pseudo three-dimensional affine transformation that matches the lifted images incorporates advantageous characteristics including spatial alignment of the surfaces, anisotropic correction of brightness/contrast differences, and stable convergence of the similarity function to its optimal value. The performance of the proposed automatic registration method is assessed against a manual method based on the projective transformation. The qualitative and quantitative assessments of the experiments have shown advantageous performance of the proposed automatic registration method against the manual one. Finally, the proposed registration method has been further improved in terms of execution time by the implementation of a surface decimation process.

In this paper, an automatic method for registering multimodal retinal images is presented. The method consists of three steps: the vessel centerline detection and extraction of bifurcation points only in the reference image, the automatic correspondence of bifurcation points in the two images using a novel implementation of the self organizing maps and the extraction of the parameters of the affine transform using the previously obtained correspondences. The proposed registration algorithm was tested on 24 multimodal retinal pairs and the obtained results show an advantageous performance in terms of accuracy with respect to the manual registration.

In this paper, a computer-aided diagnostic (CAD) system for the classification of hepatic lesions from computed tomography (CT) images is presented. Regions of interest (ROIs) taken from nonenhanced CT images of normal liver, hepatic cysts, hemangiomas, and hepatocellular carcinomas have been used as input to the system. The proposed system consists of two modules: the feature extraction and the classification modules. The feature extraction module calculates the average gray level and 48 texture characteristics, which are derived from the spatial gray-level co-occurrence matrices, obtained from the ROIs. The classifier module consists of three sequentially placed feed-forward neural networks (NNs). The first NN classifies into normal or pathological liver regions. The pathological liver regions are characterized by the second NN as cyst or "other disease." The third NN classifies "other disease" into hemangioma or hepatocellular carcinoma. Three feature selection techniques have been applied to each individual NN: the sequential forward selection, the sequential floating forward selection, and a genetic algorithm for feature selection. The comparative study of the above dimensionality reduction methods shows that genetic algorithms result in lower dimension feature vectors and improved classification performance.

Retinal image registration is commonly required in order to combine the complementary information in different retinal modalities. In this paper, a new automatic scheme to register retinal images is presented and is currently tested in a clinical environment. The scheme considers the suitability and efficiency of different image transformation models and function optimization techniques, following an initial preprocessing stage. Three different transformation models--affine, bilinear and projective--as well as three optimization techniques--downhill simplex method, simulated annealing and genetic algorithms--are investigated and compared in terms of accuracy and efficiency. The registration of 26 pairs of Fluoroscein Angiography and Indocyanine Green Chorioangiography images with the corresponding Red-Free retinal images, showed the superiority of combining genetic algorithms with the affine and bilinear transformation models. A comparative study of the proposed automatic registration scheme against the manual method, commonly used in the clinical practice, is finally presented showing the advantage of the proposed automatic scheme in terms of accuracy and consistency.