Serafeim Moustakidis - Academia.edu (original) (raw)

Papers by Serafeim Moustakidis

Aim: To test the feasibility and accuracy of a new attention-based deep learning (DL) method for ... more Aim: To test the feasibility and accuracy of a new attention-based deep learning (DL) method for right ventricular (RV) quantification using 2D echocardiography (2DE) with cardiac magnetic resonance imaging (CMR) as reference. Methods and results: We retrospectively analyzed images from 50 adult patients (median age 51, interquartile range 32-62 42% women) who had undergone CMR within 1 month of 2DE. RV planimetry of the myocardial border was performed in end-diastole (ED) and end-systole (ES) for 8 standardized 2DE RV views with calculation of areas. The DL model comprised a Feature Tokenizer module and a stack of Transformer layers. Age, gender and calculated areas were used as inputs, and the output was RV volume in ED/ES. The dataset was randomly split into training, validation and testing subsets (35, 5 and 10 patients respectively). Mean RVEDV, RVESV and RV ejection fraction (EF) were 163±70ml, 82±42ml and 51±8% respectively without differences among the subsets. The proposed ...

Information

Stroke remains a predominant cause of mortality and disability worldwide. The endeavor to diagnos... more Stroke remains a predominant cause of mortality and disability worldwide. The endeavor to diagnose stroke through biomechanical time-series data coupled with Artificial Intelligence (AI) poses a formidable challenge, especially amidst constrained participant numbers. The challenge escalates when dealing with small datasets, a common scenario in preliminary medical research. While recent advances have ushered in few-shot learning algorithms adept at handling sparse data, this paper pioneers a distinctive methodology involving a visualization-centric approach to navigating the small-data challenge in diagnosing stroke survivors based on gait-analysis-derived biomechanical data. Employing Siamese neural networks (SNNs), our method transforms a biomechanical time series into visually intuitive images, facilitating a unique analytical lens. The kinematic data encapsulated comprise a spectrum of gait metrics, including movements of the ankle, knee, hip, and center of mass in three dimensi...

Expert Systems With Applications, Mar 1, 2023

Lecture Notes in Computer Science, 2023

Zenodo (CERN European Organization for Nuclear Research), Jul 22, 2021

Explainable artificial intelligence (XAI) is an emerging new domain in which a set of processes a... more Explainable artificial intelligence (XAI) is an emerging new domain in which a set of processes and tools allow humans to better comprehend the decisions generated by black box models. However, most of the available XAI tools are often limited to simple explanations mainly quantifying the impact of individual features to the models' output. Therefore, human users are not able to understand how the features are related to each other to make predictions, whereas the inner workings of the trained models remain hidden. This paper contributes to the development of a novel graphical explainability tool that not only indicates the significant features of the model, but also reveals the conditional relationships between features and the inference capturing both the direct and indirect impact of features to the models' decision. The proposed XAI methodology, termed as gLIME, provides graphical model-agnostic explanations either at the global (for the entire dataset) or the local scale (for specific data points). It relies on a combination of local interpretable model-agnostic explanations (LIME) with graphical least absolute shrinkage and selection operator (GLASSO) producing undirected Gaussian graphical models. Regularization is adopted to shrink small partial correlation coefficients to zero providing sparser and more interpretable graphical explanations. Two well-known classification datasets (BIOPSY and OAI) were selected to confirm the superiority of gLIME over LIME in terms of both robustness and consistency/sensitivity over multiple permutations. Specifically, gLIME accomplished increased stability over the two datasets with respect to features' importance (76%-96% compared to 52%-77% using LIME). gLIME demonstrates a unique potential to extend the functionality of the current state-of-the-art in XAI by providing informative graphically given explanations that could unlock black boxes.

Clinical transplantation, Jan 18, 2021

BackgroundRecent studies demonstrated safety and efficacy of heart transplantation (HT) from hepa... more BackgroundRecent studies demonstrated safety and efficacy of heart transplantation (HT) from hepatitis C virus (HCV)‐positive donors. We sought to evaluate the impact of HCV donor status on the outcomes of patients undergoing HT in the United States.MethodsWe analyzed a retrospective cohort of adult patients from the United Network for Organ Sharing (UNOS) database who underwent isolated HT from 2015 until present. Primary outcomes were 30‐day and 1‐year overall mortality. Secondary outcomes included risk for graft failure and overall survival, incident stroke and need for dialysis during the available follow‐up period. All end points were evaluated according to HCV status.ResultsAll‐cause 30‐day and 1‐year mortality was similar between the two groups (3.4% vs 3.2%, P = .973 and 6.9% vs 7.8%, P = .769, respectively, for patients receiving heart grafts from HCV+ vs. HCV− donors). Graft failure was 12.8% (95% CI: 8%‐19%) and 15.2% (95 CI: 15%‐16%) in the HCV+ and HCV− groups, respectively (P = .92 and P = .68). Competing risk regression analysis for re‐operation showed a non‐significant trend for higher risk for re‐transplantation in the HCV+ group (HR: 2.71; 95% CI: 0.83, 8.80, P = .097).ConclusionHCV donor status does not seem to negatively affect the outcomes of HT in the U.S population.

Medical Engineering & Physics, Dec 1, 2010

A novel fuzzy decision tree-based SVM (FDT-SVM) classifier is proposed in this paper, to distingu... more A novel fuzzy decision tree-based SVM (FDT-SVM) classifier is proposed in this paper, to distinguish between asymptotic (AS) and osteoarthritis (OA) knee gait patterns and to investigate OA severity using 3-D ground reaction force (GRF) measurements. FDT-SVM incorporates effective techniques for feature selection (FS) and class grouping (CG) at each non-leaf nodes of the tree structure, which reduce the overall complexity of DT building and alleviate the overfitting effect. The embedded FS and CG are based on the notion of fuzzy partition vector (FPV) that comprises the fuzzy membership degrees of every pattern in their target classes, serving as a local evaluation metric with respect to patterns. FS is driven by a fuzzy complementary criterion (FuzCoC) which assures that features are iteratively introduced, providing the maximum additional contribution in regard to the information content given by the previously selected features. A novel Wavelet Packet (WP) decomposition based on the FuzCoC principles is also introduced, to distinguish informative and complementary features from GRF data. The quality of our method is validated in terms of statistical metrics drawn by confusion matrices, such as sensitivity, specificity and total classification accuracy. In addition, we investigate the impact of each GRF component. Finally, comparative results with existing techniques are given, demonstrating the efficacy of the suggested approach.

2022 International Joint Conference on Neural Networks (IJCNN), Jul 18, 2022

Journal of Control and Decision, Nov 6, 2022

Research Square (Research Square), Aug 10, 2021

Knee Osteoarthritis (ΚΟΑ) is a degenerative joint disease of the knee that results from the progr... more Knee Osteoarthritis (ΚΟΑ) is a degenerative joint disease of the knee that results from the progressive loss of cartilage. Due to KOA's multifactorial nature and the poor understanding of its pathophysiology, there is a need for reliable tools that will reduce diagnostic errors made by clinicians. The existence of public databases has facilitated the advent of advanced analytics in KOA research however the heterogeneity of the available data along with the observed high feature dimensionality make this diagnosis task di cult. The objective of the present study is to provide a robust Feature Selection (FS) methodology that could: (i) handle the multidimensional nature of the available datasets and (ii) alleviate the defectiveness of existing feature selection techniques towards the identi cation of important risk factors which contribute to KOA diagnosis. For this aim, we used multidisciplinary data obtained from the Osteoarthritis Initiative database for individuals without or with KOA. The proposed fuzzy ensemble feature selection methodology aggregates the results of several FS algorithms (lter, wrapper and embedded ones) based on fuzzy logic. The effectiveness of the proposed methodology was evaluated using an extensive experimental setup that involved multiple competing FS algorithms and several wellknown ML models. A 73.55 % classi cation accuracy was achieved by the best performing model (Random Forest classi er) on a group of twenty-one selected risk factors. Explainability analysis was nally performed to quantify the impact of the selected features on the model's output thus enhancing our understanding of the rationale behind the decision-making mechanism of the best model.

Medical Physics, Aug 6, 2019

Raster scan optoacoustic mesoscopy (RSOM) is evolving as a powerful alternative for non-invasive,... more Raster scan optoacoustic mesoscopy (RSOM) is evolving as a powerful alternative for non-invasive, high-resolution three-dimensional imaging of skin features based on optical absorption contrast. The technique can resolve epidermal and dermal features, including microvasculature, at resolution-to-depth ratios that go beyond optical coherence tomography (OCT) 1. For example, OCT in the visible range 2,3 can image vascular networks non-invasively to depths of only ~400 μm 4-6. As an alternative, high-frequency ultrasound can resolve microvasculature to depths of several millimeters. However, visualizing vessels with diameters smaller than 100 μm using this method requires microbubbles as contrast agents 7 , which makes it challenging to apply in humans. RSOM offers advantages of non-invasiveness and penetration depth over all these methods. For best performance, RSOM should be performed using ultra-wideband (UWB) detection, spanning a range of 200 MHz 8. UWB-RSOM can generate high-resolution images of neovascularization in a growing tumor 9 , visualize neovascularization in neoplastic gastrointestinal tissues 10 , and observe clinically relevant features of the skin microvascular structures 11-12. For this imaging technique to be clinically relevant, it is necessary to accurately define the regions and subregions of tissue in the field of view 10,13. For example, skin images should be annotated to indicate the boundaries of epidermis, dermis, and, within the dermis, the areas that have a dense microvascular structure (herein referred to as the vascular plexus), since identifying particular features in each of these subregions may facilitate disease diagnosis and assessment of its severity 8. So far, skin layers in UWB-RSOM images have been manually segmented by visual inspection of vasculature morphology or automatically based on signal intensity levels. Such procedures are slow or inaccurate and unsuitable for processing larger numbers of patients or for making clinical decisions during the patient's visit. Manual segmentation is also subjective and compromises the reproducibility and robustness of UWB-RSOM as a clinical tool.

Applied sciences, Jul 12, 2023

This article is an open access article distributed under the terms and conditions of the Creative... more This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY

Zenodo (CERN European Organization for Nuclear Research), Jun 19, 2023

Scientific Reports, Apr 24, 2023

The main goal driving this work is to develop computer-aided classification models relying on cli... more The main goal driving this work is to develop computer-aided classification models relying on clinical data to identify coronary artery disease (CAD) instances with high accuracy while incorporating the expert's opinion as input, making it a "man-in-the-loop" approach. CAD is traditionally diagnosed in a definite manner by Invasive Coronary Angiography (ICA). A dataset was created using biometric and clinical data from 571 patients (21 total features, 43% ICA-confirmed CAD instances) along with the expert's diagnostic yield. Five machine learning classification algorithms were applied to the dataset. For the selection of the best feature set for each algorithm, three different parameter selection algorithms were used. Each ML model's performance was evaluated using common metrics, and the best resulting feature set for each is presented. A stratified tenfold validation was used for the performance evaluation. This procedure was run both using the assessments of experts/doctors as input and without them. The significance of this paper lies in its innovative approach of incorporating the expert's opinion as input in the classification process, making it a "man-in-the-loop" approach. This approach not only increases the accuracy of the models but also provides an added layer of explainability and transparency, allowing for greater trust and confidence in the results. Maximum achievable accuracy, sensitivity, and specificity are 83.02%, 90.32%, and 85.49% when using the expert's diagnosis as input, compared to 78.29%, 76.61%, and 86.07% without the expert's diagnosis. The results of this study demonstrate the potential for this approach to improve the diagnosis of CAD and highlight the importance of considering the role of human expertise in the development of computer-aided classification models. In modern society, cardiovascular diseases (CVD) are one of the most common health problems and the leading cause of death 1. According to the World Health Organization (WHO), an estimated 17.9 million people died from CVDs in 2019, representing 32% of all global deaths 2. Coronary Artery Disease (CAD) is the most common type of cardiovascular disease. It is thus safe to assume that timely diagnosis and treatment of CAD can lead to a significant decrease in global mortality. This work aims to introduce a new tool based on Machine Learning (ML) classification models meant to be used in a consultative manner by experts in the process of diagnosis. Single Photon Emission Computed Tomography (SPECT) and Positron Emission Tomography (PET) are the main tools used for the visual assessment of CAD. Experts visually interpret these images and, by combining clinical, historical, and biometric data, conclude on a verdict and its consequent treatment. On the other hand, recently, Artificial Intelligence (AI) has significantly supported healthcare intervention decisions via customized Medical Decision Support Systems (MDSS). A repertoire of predictive models based on boosted trees, random forests, and straightforward Deep Learning (DL) have provided highly accurate predictions for multi-parameter and complex designs in medicine 3. Imaging technologies and clinical data do not ensure a definite diagnosis, whilst there are many situations that pose diagnostic dilemmas. Therefore, it is very often that a patient is

Computer Methods in Biomechanics and Biomedical Engineering, Jun 1, 2012

An efficient wavelet-based feature selection (FS) method is proposed in this paper for subject re... more An efficient wavelet-based feature selection (FS) method is proposed in this paper for subject recognition using ground reaction force measurements. Our approach relies on a local fuzzy evaluation measure with respect to patterns that reveal the adequacy of data coverage for each feature. Furthermore, FS is driven by a fuzzy complementary criterion (FuzCoC) which assures that those features are iteratively introduced, providing the maximum additional contribution with regard to the information content given by the previously selected features. On the basis of the principles of FuzCoC, we develop two novel techniques. At Stage 1, wavelet packet (WP) decomposition of gaits is accomplished to obtain a set of discriminating frequency sub-bands. A computationally simple FS method is then applied at Stage 2, providing a compact set of powerful and complementary features, from WP coefficients. The quality of our approach is validated via comparative analysis against existing methods on gait recognition.

Every year, the rate at which technology is applied on areas of our everyday life is increasing a... more Every year, the rate at which technology is applied on areas of our everyday life is increasing at a steady pace. This rapid development drives the technology companies to design and fabricate their integrated circuits (ICs) in non-trustworthy outsourcing foundries in order to reduce the cost. Thus, a synchronous form of virus, known as Hardware Trojans (HTs), was developed. HTs leak encrypted information, degrade device performance or lead to total destruction. To reduce the risks associated with these viruses, various approaches have been developed aiming to prevent and detect them, based on conventional or machine learning methods. Ideally, any undesired modification made to an IC should be detectable by pre-silicon verification/simulation and post-silicon testing. The infected circuit can be inserted in different stages of the manufacturing process, rendering the detection of HTs a complicated procedure. In this paper, we present a comprehensive review of research dedicated to applications based on Machine Learning for the detection of HTs in ICs. The literature is categorized in (a) reverse-engineering development for the imaging phase, (b) real-time detection, (c) golden model-free approaches, (d) detection based on gate-level netlists features and (e) classification approaches.

BioMed, Dec 27, 2022

Computational Materials Science, Mar 1, 2023