Dimitrios Hristu-Varsakelis

𝐵𝑎𝑐𝑘𝑔𝑟𝑜𝑢𝑛𝑑𝑎𝑛𝑑𝑂𝑏𝑗𝑒𝑐𝑡𝑖𝑣𝑒: Paroxysmal atrial fibrillation (PAF) is a transient form of AF that can be asymptomatic, with no visible signs on the electrocardiogram (ECG). PAF can lead to stroke and thrombotic events, thus its timely detection and therapeutic approach are crucial. The aim of this study was to develop an explainable machine learning-based model which achieves state-of-the-art PAF identification during sinus rhythm by assessing atrial activity and conduction using three-dimensional vectrorcardiographic (VCG) indices in an asymptomatic, at-risk population. Methods: We examined 59 healthy subjects and 69 subjects with a history of PAF and constructed theirP-wave loops from 10-minute VCG recordings during sinus rhythm, in order to assess beat-to-beat the atrial electrical activity. We computed various dynamic and static features of the loops, including the eigenvalues, the geometric characterization of the P-loop vector, and quantification indices of P-wave morphology. We found that the majority of these features presented significant statistical differences between the two groups. We applied a Random Forest classifier with recursive feature elimination using 10-fold cross-validation. Results: We observed that the eigenvalues of the P-loop projection onto the sagittal and transverse planes, along with the area of the loop-fitted ellipsoid, and the average energy of P-loop vector were sufficient to correctly classify subjects with a state-of-the-art accuracy ACC=95.23%. 𝐶𝑜𝑛𝑐𝑙𝑢𝑠𝑖𝑜𝑛𝑠: Based on the three-dimensional representation of atrial activity encoded in the P-waveloops, we propose clinically significant indices that are visually interpretable, render the classifier results explainable, and are clinically relevant to physicians. Our classifier could be incorporated into clinical practice as a reliable and non-black box decision guide identifying subjects at AF risk.

One of the crucial indicators of the heart's functioning, is the so-called left ventricular ejection fraction (LVEF), which measures the heart's ability to pump blood, and corresponds to the relative change in volume within the heart's left ventricle between it's most expanded (end-diastole) and most contracted state (end-systole) during a cardiac cycle. A reduced LVEF is a key indicator of heart failure, and as such, its accurate measurement plays a prominent role in cardiology. This work proposes a machine learning approach for estimating the LVEF from short echocardiogram videos. Our model, based on gradient-boosted trees, is significantly simpler than the state of the art, but is competitive in terms of accuracy and has a higher degree of explainability. The proposed model operates on a set of geometric features of the heart's left ventricle, tracking its evolution during the cardiac cycle; some of these features are novel and are proposed here for the first time. We discuss the performance of our model on a dataset of over 10,000 samples, including the relative importance of our proposed features, and show that the model's estimation error is well within the margin of variation that occurs when the same LVEF is measured by different experts.

Blind signature schemes (BSS) play a pivotal role in privacy‐oriented cryptography. However, with BSS, the signed message remains unintelligible to the signer, giving them no guarantee that the blinded message he signed actually contained valid information. Partially BSS (PBSS) were introduced to address precisely this problem. In this study, the authors present the first leakage‐resilient, lattice‐based PBSS in the literature. The proposed construction is provably secure in the random oracle model and offers quasi‐linear complexity w.r.t. key/signature sizes and signing speed. In addition, it offers statistical partial blindness and its unforgeability is based on the computational hardness of worst‐case ideal lattice problems for approximation factors in in dimension n . The proposed scheme benefits from the subexponential hardness of ideal lattice problems and remains secure even if a fraction of the signer's secret key leaks to an adversary via arbitrary side‐channels. Several extensions of the security model, such as honest‐user unforgeability and selective failure blindness, are also considered and concrete parameters for instantiation are proposed.

We present a deep long short-term memory (LSTM)-based neural network for predicting asset prices, together with a successful trading strategy for generating profits based on the model's predictions. Our work is motivated by the fact that the effectiveness of any prediction model is inherently coupled to the trading strategy it is used with, and vise versa. This highlights the difficulty in developing models and strategies which are jointly optimal, but also points to avenues of investigation which are broader than prevailing approaches. Our LSTM model is structurally simple and generates predictions based on price observations over a modest number of past trading days. The model's architecture is tuned to promote profitability, as opposed to accuracy, under a strategy that does not trade simply based on whether the price is predicted to rise or fall, but rather takes advantage of the distribution of predicted returns, and the fact that a prediction's position within that distribution carries useful information about the expected profitability of a trade. The proposed model and trading strategy were tested on the S&P 500, Dow Jones Industrial Average (DJIA), NASDAQ and Russel 2000 stock indices, and achieved cumulative returns of 340%, 185%, 371% and 360%, respectively, over 2010-2018, far outperforming the benchmark buy-and-hold strategy as well as other recent efforts.