Literatura académica sobre el tema "Modelli diagnostici ML"

Short of a liver biopsy, hepatic disease and drug-induced liver injury are diagnosed and classified from clinical findings, especially laboratory results. It was hypothesized that a healthy hepatic dynamic equilibrium might be modelled by an Ornstein–Uhlenbeck (OU) stochastic process, which might lead to more sensitive and specific diagnostic criteria. Using pooled data from healthy volunteers in pharmaceutical clinical trials, this model was applied using maximum likelihood (ML) methods. It was found that the exponent of the autocorrelation function was proportional to the square root of time rather than time itself, as predicted by the OU model. This finding suggests a stronger autocorrelation than expected and may have important implications regarding the use of laboratory testing in clinical diagnosis, in clinical trial design, and in monitoring drug safety. Besides rejecting the OU hypothesis for liver test homeostasis, this paper presents ML estimates for the multivariate Gaussian distribution for healthy adult males. This work forms the basis for a new approach to mathematical modelling to improve both the sensitivity and specificity of clinical measurements over time.

The term artificial intelligence (AI) was proposed in 1956 by Dartmouth scholar John McCarthy, which refers to hardware or software that exhibits behavior which appears intelligent.1 During recent times, AI gained immense popularity as new algorithms, specialized hardware, huge data and cloud-based services were developed. Machine learning (ML), a subset of AI, originated in 1980 and is defined as a set of methods that automatically detect patterns in data and then incorporate this information to predict future data under uncertain conditions. Another escalating technology of ML called Deep learning (DL), launched in 2000s, is an escalating technology of ML and has revolutionized the world of AI. These technologies are powerful tools utilized by modern society for objects' recognition in images, real-time languages' translation, device manipulation via speech (such as Apple's Siri®, Amazon’s Alexa®, Microsoft’s Cortana®, etc.). The steps for AI model include preprocessing image data, train, validate and test the model, and evaluate the trained model's performance. To increase AI prediction efficiency, raw data need to be preprocessed. Data collected from different sources needs to be integrated and the most relevant features selected and extracted to improve the learning process performance. Data set is randomly partitioned into two independent subsets, one is for modeling and the other is for testing. The test set is used to evaluate the final performance of the trained model. The area under receiver operating characteristic curves (AUC) is most used evaluation metrics for quantitative assessment of a model in AI diagnosis. The AUCs effective models range from 0.5 to 1; higher the value of AUC, better the performance of the model.2 In the medical field, AI gained popularity by visualization of input images of highly potential abnormal sites which can be reviewed and analyzed in future.
 AI and DL algorithms or systems are also widely used in field of ophthalmology. More intensively studied fields are diabetic retinopathy, age related macular degeneration, and cataract and glaucoma. Various ophthalmic imaging modalities used for AI diagnosis include fundus image, optical coherence tomography (OCT), ocular ultrasound, slit-lamp image and visual field. Diabetic retinopathy (DR), a diabetic complication, is a vasculopathy that affects one-third of diabetic patients leading to irreversible blindness. AI has been in use to predict DR risk and its progression. Gulshan and colleague were the first to report the application of DL for DR identification.3 They used large fundus image data sets in supervised manner for DR detection. Other studies applied DL to identify and stage DR. DL-based computer-aided system was introduced to detect DR through OCT images, achieving a specificity of 0.98.4 A computer-aided diagnostic (CAD) system based on CML algorithms using optical coherence tomography angiography images to automatically diagnose non-proliferative DR (NPDR) also achieved high accuracy and AUC.5 Age-related macular degeneration (AMD) is the leading cause of irreversible blindness among old people in the developed world. ML algorithms are being used to identify AMD lesions and prompt early treatment with accuracy usually over 80%.6 Using ML to predict treatment of retinal neovascularity in AMD and DR by anti-vascular endothelial growth factor (Anti VEGF) injection requirements can manage patients' economic burden and resource management. ML algorithms have been applied to diagnose and grade cataract using fundus images, ultrasounds images, and visible wavelength eye images.7 Glaucoma is the third largest sight-threatening eye disease around the world. Glaucoma patients suffered from high intraocular pressure, damage of the optic nerve head, retina nerve fiber layer defect, and gradual vision loss. Studies using DL methods to diagnose glaucoma are few. So far, fundus images and wide-field OCT scans have all been used to construct DL-based glaucomatous diagnostic models. Mostly, the DL-based methods show excellent results.8
 In this era of “evidence-based medicine,” clinicians and patients find it difficult to trust a mysterious machine to diagnose yet cannot provide explanations of why the patient has certain disease. In future, advanced AI interpreters will be launched which will contribute significantly to revolutionize current disease diagnostic pattern.

Il ruolo fondamentale che la manutenzione gioca nei costi di esercizio e nella produttività degli impianti industriali ha portato le aziende e i ricercatori a spostare il loro interesse su questo tema. L'ultima frontiera dell'innovazione in campo manutentivo, resa possibile anche dall'avvento della quarta rivoluzione industriale che promuove la sensorizzazione e l’interconnessione di tutti i macchinari di impianto, è la manutenzione predittiva. Essa mira ad ottenere una previsione accurata della vita utile dei componenti degli impianti industriali al fine di ottimizzare la schedulazione degli interventi sul campo. Lo studio parte da una accurata revisione della letteratura scientifica di settore riguardante le tecniche diagnostiche e prognostiche applicate a componenti di impianti industriali, necessaria alla comprensione dei diversi modelli sviluppati in funzione della tipologia di componente e modo di guasto in analisi. Successivamente ho spostato l’attenzione sul concetto di manutenzione 4.0 al fine di mappare tutte le caratteristiche associate al paradigma dell'Industria 4.0 e le loro possibili applicazioni alla manutenzione. Lo studio condotto ha portato poi alla progettazione, sviluppo e validazione delle metodologie necessarie all’applicazione in real-time di modelli diagnostici e prognostici avanzati, sia statistici che machine learning, necessari all’implementazione sul campo di un sistema di manutenzione predittiva. Grazie all’applicazione delle metodologie proposte ad un caso studio è stato possibile non solo validare i modelli proposti ma anche definire l’architettura informatica necessaria alla loro corretta implementazione sul sistema distribuito di controllo (Distributed Control System - DCS) di impianto in funzione della tipologia del componente e del guasto in analisi. I modelli testati e validati hanno mostrato elevate prestazioni diagnostiche soprattutto per quanto riguarda i modelli ML che sfruttano le Support Vector Machine (SVM). In definitiva, questo lavoro di tesi mostra nel dettaglio tutti i passaggi necessari allo sviluppo di un sistema di manutenzione predittiva efficace in impianto: partendo dall’analisi dei modi di guasto e dalla sensorizzazione dei componenti, passando poi allo sviluppo dei modelli diagnostici e prognostici real-time fino alla costruzione dell’interfaccia di visualizzazione dei risultati delle analisi svolte, analizzando anche l’architettura informatica necessaria al suo corretto funzionamento. The fundamental role that maintenance plays in the operating costs and productivity of industrial plants has led companies and researchers to shift their interest in this issue. The last frontier of innovation in the maintenance field, made possible also by the advent of the fourth industrial revolution which promotes the sensorisation and interconnection of all plant machinery, is predictive maintenance. It aims to obtain an accurate forecast of the useful life of the industrial plants’ components in order to optimise the scheduling of interventions in the field. The study starts from an accurate review of the scientific literature concerning the diagnostic and prognostic techniques applied to industrial plant components, necessary to understand the different models developed according to the type of component and failure mode under analysis. Subsequently I shifted the focus to the maintenance 4.0 concept in order to map all the characteristics associated with the Industry 4.0 paradigm and their possible applications to maintenance operations. The study then led to the design, development and validation of the methodologies necessary for the real-time application of advanced diagnostic and prognostic models, both statistical and machine learning, necessary for the field implementation of a predictive maintenance system. Thanks to the application of the proposed methodologies to a case study, it was possible not only to validate the proposed models but also to define the IT architecture necessary for their correct implementation on the plant's Distributed Control System (DCS) according to the type of component and the fault under analysis. The tested and validated models showed high diagnostic performance, especially regarding the Support Vector Machine (SVM) Machine Learning models. Ultimately, this thesis shows in detail all the steps necessary for the development of an effective predictive maintenance system in the plant: starting from the analysis of failure modes and component sensorisation, then moving on to the development of real-time diagnostic and prognostic models up to the build-up of the interface for visualising the results of the analyses carried out, also analysing the IT architecture necessary for its correct operation.

The growing success of Machine Learning (ML) is making significant improvements to predictive models, facilitating their integration in various application fields, especially the healthcare context. However, it still has limitations and drawbacks, such as the lack of interpretability which does not allow users to understand how certain decisions are made. This drawback is identified with the term "Black-Box", as well as models that do not allow to interpret the internal work of certain ML techniques, thus discouraging their use. In a highly regulated and risk-averse context such as healthcare, although "trust" is not synonymous with decision and adoption, trusting an ML model is essential for its adoption. Many clinicians and health researchers feel uncomfortable with black box ML models, even if they achieve high degrees of diagnostic or prognostic accuracy. Therefore more and more research is being conducted on the functioning of these models. Our study focuses on the Random Forest (RF) model. It is one of the most performing and used methodologies in the context of ML approaches, in all fields of research from hard sciences to humanities. In the health context and in the evaluation of health policies, their use is limited by the impossibility of obtaining an interpretation of the causal links between predictors and response. This explains why we need to develop new techniques, tools, and approaches for reconstructing the causal relationships and interactions between predictors and response used in a RF model. Our research aims to perform a machine learning experiment on several medical datasets through a comparison between two methodologies, which are inTrees and NodeHarvest. They are the main approaches in the rules extraction framework. The contribution of our study is to identify, among the approaches to rule extraction, the best proposal for suggesting the appropriate choice to decision-makers in the health domain.

Recently, disease prediction using diagnostic reports and images are one of the most popular applications of artificial intelligence (AI) and machine learning (ML). Several authors reported significant results in this area by combining cutting-edge hardware with AI and ML-based technologies. In this chapter, the authors present a review of different works carried for the prediction of several chronic diseases by researchers in last five years. Reported AI and ML based methodologies have been used to forecast chronic disease such as heart problems, brain tumors, asthma, diabetes, cholera, arthritis, liver diseases, kidney diseases, malaria, and leukemia. In the literature, the authors also discuss the different user interfaces which have been used to interact with real time AI and ML based disease prediction models. The authors have presented the detailed discussion of each paper including advantages, disadvantages, datasets, performance metrics such as precision, recall, accuracy and F1 score. In the final section, the survey concludes with a description of research gaps that can be addressed by future research attempts.

This paper aims to explore physicians’ adoption of Machine learning models in the healthcare process and barriers that may hinder it. A review of the literature about ML in healthcare included current and potentially beneficial clinical applications and clinicians’ adoption and trust towards such applications. While some physicians are looking forward to using ML to improve their outcomes and reduce their load, we uncovered fear of unwanted outcomes and concerns about privacy of data, legal liability, and patient dissatisfaction.

One of the most important tasks in database technology is to combine the following activities: data mining or inferring knowledge from data and query processing or reasoning on acquired knowledge. The solution of this task requires a logical language with unified syntax and semantics for integrating deductive (using knowledge) and inductive (acquiring knowledge) reasoning. In this paper, we propose a unified model of commonsense reasoning. We also demonstrate that a large class of inductive machine learning (ML) algorithms can be transformed into the commonsense reasoning processes based on wellknown deduction and induction logical rules. The concept of a good classification (diagnostic) test (Naidenova & Polegaeva, 1986) is the basis of our approach to combining deductive and inductive reasoning.

One of the most important tasks in database technology is to combine the following activities: data mining or inferring knowledge from data and query processing or reasoning on acquired knowledge. The solution of this task requires a logical language with unified syntax and semantics for integrating deductive (using knowledge) and inductive (acquiring knowledge) reasoning. In this paper, we propose a unified model of commonsense reasoning. We also demonstrate that a large class of inductive machine learning (ML) algorithms can be transformed into the commonsense reasoning processes based on wellknown deduction and induction logical rules. The concept of a good classification (diagnostic) test (Naidenova & Polegaeva, 1986) is the basis of our approach to combining deductive and inductive reasoning.

Societal evolution has resulted in a complex lifestyle where we give most attention to our physical health leaving psychological health less prioritized. Considering the complex relationship between stress and psychological well-being, this study bases itself on the cognitive states experienced by us. The presented research offers insight into how state-of-the-art technologies can be used to support positive cognitive states. It makes use of the brain-computer interface (BCI) that drives the data collection using electroencephalography (EEG). The study leverages data science to devise machine learning (ML) model to predict the corresponding stress levels of an individual. A feedback loop using “Self Quantification” and “Nudging” offer real-time insights about an individual. Such a mechanism can also support the psychological conditioning of an individual where it does not only offer spatial flexibility and cognitive assistance but also results in enhanced self-efficacy. Being part of quantified self-movement, such an experimental approach could showcase personalized indicators to reflect a positive cognitive state. Although ML modeling in such a data-driven approach might experience reduced diagnostic sensitivity and suffer from observer variability, it can complement psychosomatic treatments for preventive healthcare.

Diabetes is one of the most deadly diseases on the planet. It is also a cause of a variety of illnesses, such as coronary artery disease, blindness, and urinary organ disease. In this situation, the patient must visit a medical center to obtain their results following consultation. Finding the right combination of characteristics and machine learning techniques for classification is also very critical. However, with the advancement of machine learning techniques, we now have the potential to find a solution to the current problem. The healthcare recommendation system (HRS) may be designed to predict health by evaluating patient lifestyle, physical health, mental health aspects using machine learning. For example, training the model using people's age and diabetes helps to predict new patients without a specific diagnostic for diabetes. The proposed deep learning model with convolutional neural network (D-CNN) achieves an overall accuracy of 96.25%. D-CNN is found to be more successful for diabetes prediction than other machine learning (ML) approaches in the experimental analysis.

Endometriosis is a commonly occurring progressive gynecological disorder, in which tissues similar to the lining of the uterus grow on other parts of the female body, including ovaries, fallopian tubes, and bowel. It is one of the primary causes of pelvic discomfort and fertility challenges in women. The actual cause of the endometriosis is still undetermined. As a result, the objective of the chapter is to identify the drivers of endometriosis’ diagnoses via leveraging selected advanced machine learning (ML) algorithms. The primary risks of infertility and other health complications can be minimized to a greater extent if a likelihood of endometriosis could be predicted well in advance. Logistic regression (LR) and eXtreme Gradient Boosting (XGB) algorithms leveraged 36 months of medical history data to demonstrate the feasibility. Several direct and indirect features were identified as important to an accurate prediction of the condition onset, including selected diagnosis and procedure codes. Creating analytical tools based on the model results that could be integrated into the Electronic Health Records (EHR) systems and easily accessed by healthcare providers might aid the objective of improving the diagnostic processes and result in a timely and precise diagnosis, ultimately increasing patient care and quality of life.

Abstract Numerical simulation results are the basis of numerous oil and gas field developments. We based the numerical simulation models (or dynamic models) on 3D geological models. We constructed a geological model using core and log data obtained from wells as inputs to create a reservoir prototype. This paper describes the applications of artificial intelligence (AI) algorithms for parameterization of static and dynamic modeling processes. Accordingly, a hypothetical 3D geological model was created, and porosity and permeability were distributed using sequential Gaussian simulation. Then, Petro-physical rock types (PRT) were defined in the 3D space as a function of porosity and permeability using a hypothetical Winland's R35 equation. Finally, hypothetical saturation-height functions (SHFs) were defined for different PRTs to populate water saturation in the 3D geological model. Subsequently, some wells were randomly defined in the 3D model to obtain the logs of porosity, permeability, SHF, PRT, repeat formation tester pressure (RFT), and datum pressures that are used in this study. A multivariate Gaussian regression was applied for anomaly detection, while core porosity and permeability were filtered. Subsequently, a fixed window average was used to detect the boundaries of core data stationarity and propose the optimum reservoir zone required to describe the internal heterogeneities of the reservoir. Then, we deployed the k-means clustering algorithm to determine the PRT and saturation height function (SHF) based on the core and log data derived from the hypothetical geological model. Finally, we used the clustering-based pattern recognition to cluster well datum pressures into homogeneous groups and create a connected reservoir region CRR map to be used as an input in the 3D permeability distribution. Our results demonstrate the value of additional diagnostics that can be used in conjunction with the traditional semi-log plot of porosity and permeability. This additional diagnostic approach is a semi-log plot of permeability versus depth, which can help check whether intra-reservoir heterogeneities observable in core data have been preserved in the 3D model. In our case, a 3D model created using the core and log data from the hypothetical model and honoring the internal reservoir architecture resulted in a better history match regarding the hypothetical geo-model's RFT pressure signature. Our results further demonstrate that PRT and SHF derived from k-means clustering are sufficiently similar to those of the hypothetical model. Time series anomaly filtering of pressures helped detect incorrect well data that may otherwise have gone unnoticed. Using the nearest-neighbor property distribution resulted in a geological model whose diagnostic plots indicated an excellent match with core data and allowed a better assessment of modeling uncertainties. The ML approaches presented in this study could help obtain data-derived PRT and SHF to complement Winland's interpretation when Mercury Injection Capillary Pressure (MICP) experiments are limited or unavailable, saving both time and cost. Using the fixed window averaging helps optimize the geological model zone assessment, resulting in a better intra-reservoir architecture. Finally, we derive insights into a more efficient core acquisition plan.

Objectives 1. Establish diagnostic procedures to identify tolerant carrier birds based on a) Isolation of ALV-J from blood, b) Detection of group-specific antigen in cloacal swabs and egg albumen. Application of these procedures to broiler breeder flocks with the purpose of removing virus positive birds from the breeding program. 2. Survey the AL V-J infection status of foundation lines to estimate the feasibility of the eradication program 3. Investigate virus transmission through the embryonated egg (vertical) and between chicks in the early post-hatch period (horizontal). Establish a model for limiting horizontal spread by analyzing parameters operative in the hatchery and brooder house. 4. Compare the pathogenicity of AL V-J isolates for broiler chickens. 5. Determine whether AL V-J poses a human health hazard by examining its replication in mammalian and human cells. Revisions. The: eradication objective had to be terminated in the second year following the closing down of the Poultry Breeders Union (PBU) in Israel. This meant that their foundation flocks ceased to be available for selection. Instead, the following topics were investigated: a) Comparison of commercial breeding flocks with and without myeloid leukosis (matched controls) for viremia and serum antibody levels. b) Pathogenicity of Israeli isolates for turkey poults. c) Improvement of a diagnostic ELISA kit for measuring ALV-J antibodies Background. ALV-J, a novel subgroup of the avian leukosis virus family, was first isolated in 1988 from broiler breeders presenting myeloid leukosis (ML). The extent of its spread among commercial breeding flocks was not appreciated until the disease appeared in the USA in 1994 when it affected several major breeding companies almost simultaneously. In Israel, ML was diagnosed in 1996 and was traced to grandparent flocks imported in 1994-5, and by 1997-8, ML was present in one third of the commercial breeding flocks It was then realized that ALV-J transmission was following a similar pattern to that of other exogenous ALVs but because of its unusual genetic composition, the virus was able to establish an extended tolerant state in infected birds. Although losses from ML in affected flocks were somewhat higher than normal, both immunosuppression and depressed growth rates were encountered in affected broiler flocks and affected their profitability. Conclusions. As a result of the contraction in the number of international primary broiler breeders and exchange of male and female lines among them, ALV-J contamination of broiler breeder flocks affected the broiler industry worldwide within a short time span. The Israeli national breeding company (PBU) played out this scenario and presented us with an opportunity to apply existing information to contain the virus. This BARD project, based on the Israeli experience and with the aid of the ADOL collaborative effort, has managed to offer solutions for identifying and eliminating infected birds based on exhaustive virological and serological tests. The analysis of factors that determine the efficiency of horizontal transmission of virus in the hatchery resulted in the workable solution of raising young chicks in small groups through the brooder period. These results were made available to primary breeders as a strategy for reducing viral transmission. Based on phylogenetic analysis of selected Israeli ALV-J isolates, these could be divided into two groups that reflected the countries of origin of the grandparent stock. Implications. The availability of a simple and reliable means of screening day old chicks for vertical transmission is highly desirable in countries that rely on imported breeding stock for their broiler industry. The possibility that AL V-J may be transmitted to human consumers of broiler meat was discounted experimentally.