Trustworthy Precision Medicine: An Interpretable Approach to Detecting Anomalous Behavior of IoT Devices.

The growing integration of Internet of Things (IoT) technology within the healthcare sector has revolutionized healthcare delivery, enabling advanced personalized care and precise treatments. However, this raises significant challenges, demanding robust, intelligible, and effective monitoring mechanisms. We propose an interpretable machine-learning approach to the trustworthy and effective detection of behavioral anomalies within the realm of medical IoT. The discovered anomalies serve as indicators of potential system failures and security threats. Essentially, the detection of anomalies is accomplished by learning a classifier from the operational data generated by smart devices. The learning problem is dealt with in predictive association modeling, whose expressiveness and intelligibility enforce trustworthiness to offer a comprehensive, fully interpretable, and effective monitoring solution for the medical IoT ecosystem. Preliminary results show the effectiveness of our approach.

AI Approach for Enhanced Thalassemia Diagnosis Using Blood Smear Images.

This paper aims to propose an approach leveraging Artificial Intelligence (AI) to diagnose thalassemia through medical imaging. The idea is to employ a U-net neural network architecture for precise erythrocyte morphology detection and classification in thalassemia diagnosis. This accomplishment was realized by developing and assessing a supervised semantic segmentation model of blood smear images, coupled with the deployment of various data engineering techniques. This methodology enables new applications in tailored medical interventions and contributes to the evolution of AI within precision healthcare, establishing new benchmarks in personalized treatment planning and disease management.

Enhancing Thrombophilia Risk Prediction Through AI-Based Methodologies.

Thrombophilia, a predisposition to thrombosis, poses significant diagnostic challenges due to its multi-factorial nature, encompassing genetic and acquired factors. Current diagnostic paradigms, primarily relying on a combination of clinical assessment and targeted laboratory tests, often fail to capture the complex interplay of factors contributing to thrombophilia risk. This paper proposes an innovative artificial intelligence (AI)-based methodology aimed to enhance the prediction of thrombophilia risk. The designed multidimensional risk assessment model integrates and elaborates through AI a comprehensive collection of patient data types, including genetic markers, clinical parameters, patient history, and lifestyle factors, in order to obtain advanced and personalized explainable diagnoses.

Swarm Intelligence to Face IoT Challenges.

The Internet of Things (IoT) paradigm denotes billions of physical entities connected to Internet that allow the collecting and sharing of big amounts of data. Everything may become a component of the IoT thanks to advancements in hardware, software, and wireless network availability. Devices get an advanced level of digital intelligence that enables them to transmit real-time data without applying for human support. However, IoT also comes with its own set of unique challenges. Heavy network traffic is generated in the IoT environment for transmitting data. Reducing network traffic by determining the shortest route from the source to the aim decreases overall system response time and energy consumption costs. This translates into the need to define efficient routing algorithms. Many IoT devices are powered by batteries with limited lifetime, so in order to ensure remote, continuous, distributed, and decentralized control and self-organization of these devices, power-aware techniques are highly desirable. Another requirement is to manage huge amounts of dynamically changing data. This paper reviews a set of swarm intelligence (SI) algorithms applied to the main challenges introduced by the IoT. SI algorithms try to determine the best path for insects by modeling the hunting behavior of the agent community. These algorithms are suitable for IoT needs because of their flexibility, resilience, dissemination degree, and extension.

Low Cost and Easy Validation Anticounterfeiting Plasmonic Tags Based on Thin Films of Metal and Dielectric.

Multilevel anticounterfeiting Physical Unclonable Function (PUF) tags based on thin film of silver (Ag), Zinc Oxide (ZnO) and PolyVinylPyrrolidone (PVP), are experimentally demonstrated and validated. We exploit the low adhesion of silver to glass and consequent degradation during ZnO deposition to induce morphological randomness. Several photographs of the tag surfaces have been collected with different illumination conditions and using two smartphones of diverse brand. The photos were analyzed using an image recognition algorithm revealing low common minutiae for different tags. Moreover, the optical response reveals peculiar spectra due to labels of plasmonic nature. The proposed systems can be easily fabricated on large areas and represent a cost-effective solution for practical protection of objects.

Kidney Tumor Detection and Classification Based on Deep Learning Approaches: A New Dataset in CT Scans.

Kidney tumor (KT) is one of the diseases that have affected our society and is the seventh most common tumor in both men and women worldwide. The early detection of KT has significant benefits in reducing death rates, producing preventive measures that reduce effects, and overcoming the tumor. Compared to the tedious and time-consuming traditional diagnosis, automatic detection algorithms of deep learning (DL) can save diagnosis time, improve test accuracy, reduce costs, and reduce the radiologist's workload. In this paper, we present detection models for diagnosing the presence of KTs in computed tomography (CT) scans. Toward detecting and classifying KT, we proposed 2D-CNN models; three models are concerning KT detection such as a 2D convolutional neural network with six layers (CNN-6), a ResNet50 with 50 layers, and a VGG16 with 16 layers. The last model is for KT classification as a 2D convolutional neural network with four layers (CNN-4). In addition, a novel dataset from the King Abdullah University Hospital (KAUH) has been collected that consists of 8,400 images of 120 adult patients who have performed CT scans for suspected kidney masses. The dataset was divided into 80% for the training set and 20% for the testing set. The accuracy results for the detection models of 2D CNN-6 and ResNet50 reached 97%, 96%, and 60%, respectively. At the same time, the accuracy results for the classification model of the 2D CNN-4 reached 92%. Our novel models achieved promising results; they enhance the diagnosis of patient conditions with high accuracy, reducing radiologist's workload and providing them with a tool that can automatically assess the condition of the kidneys, reducing the risk of misdiagnosis. Furthermore, increasing the quality of healthcare service and early detection can change the disease's track and preserve the patient's life.

Intrusion Detection System for IoT Based on Deep Learning and Modified Reptile Search Algorithm.

This study proposes a novel framework to improve intrusion detection system (IDS) performance based on the data collected from the Internet of things (IoT) environments. The developed framework relies on deep learning and metaheuristic (MH) optimization algorithms to perform feature extraction and selection. A simple yet effective convolutional neural network (CNN) is implemented as the core feature extractor of the framework to learn better and more relevant representations of the input data in a lower-dimensional space. A new feature selection mechanism is proposed based on a recently developed MH method, called Reptile Search Algorithm (RSA), which is inspired by the hunting behaviors of the crocodiles. The RSA boosts the IDS system performance by selecting only the most important features (an optimal subset of features) from the extracted features using the CNN model. Several datasets, including KDDCup-99, NSL-KDD, CICIDS-2017, and BoT-IoT, were used to assess the IDS system performance. The proposed framework achieved competitive performance in classification metrics compared to other well-known optimization methods applied for feature selection problems.

A Microarray Analysis Technique Using a Self-Organizing Multiagent Approach.

Microarray technology is fully established among the research fields in genetic domain. Academia and industrial researchers investigate and analyze genes' expression to obtain more and more useful information about given organisms, with the aim to perform better disease diagnosis and prediction, accurate medical data analysis, etc. Analyzing gene expression data, often available in raw form, implies a huge amount of analytical and computational complexities and therefore, innovative and intelligent mechanisms have to be designed to obtain useful information from this precious data. This chapter proposes a multiagent algorithm for building a distributed algorithm for DNA Microarray management. A collection of agents, in which each one representing a Microarray (or chip), execute in parallel a sequence of simple operations exploiting local information, and an organized virtual structure is built at global level. A word embeddings approach, able to capture the semantic context and represent Microarrays with vectors, is employed to map the chips, so allowing advanced agents' operations. A similarity-based overlay network of agents is brought out and an efficient management system of DNA Microarray is enabled. The generated virtual structure allows executing of informed operations, such as range queries, in a large dataset containing unstructured data. Preliminary results were confirm the validity of the algorithm proposed.

Hybrid Plasmonic/Photonic Nanoscale Strategy for Multilevel Anticounterfeit Labels.

Innovative goods authentication strategies are of fundamental importance considering the increasing counterfeiting levels. Such a task has been effectively addressed with the so-called physical unclonable functions (PUFs), being physical properties of a system that characterize it univocally. PUFs are commonly implemented by exploiting naturally occurring non-idealities in clean-room fabrication processes. The broad availability of classic paradigm PUFs, however, makes them vulnerable. Here, we propose a hybrid plasmonic/photonic multilayered structure working as a three-level strong PUF. Our approach leverages on the combination of a functional nanostructured surface, a resonant response, and a unique chromatic signature all together in one single device. The structure consists of a resonant cavity, where the top mirror is replaced with a layer of plasmonic Ag nanoislands. The naturally random spatial distribution of clusters and nanoparticles formed by this deposition technique constitutes the manufacturer-resistant nanoscale morphological fingerprint of the proposed PUF. The presence of Ag nanoislands allows us to tailor the interplay between the photonic and plasmonic modes to achieve two additional security levels. The first one is constituted by the chromatic response and broad iridescence of our structures, while the second by their rich spectral response, accessible even through a common smartphone light-emitting diode. We demonstrate that the proposed architectures could also be used as an irreversible and quantitative temperature exposure label. The proposed PUFs are inexpensive, chip-to-wafer-size scalable, and can be deposited over a variety of substrates. They also hold a great promise as an encryption framework envisioning morpho-cryptography applications.