Tesi di Dottorato
Permanent URI for this communityTesi di Dottorato
Browse
3 results
Search Results
Item Towards an effective and explainable AI: studies in the biomedical domain(Università della Calabria, 2021-07-05) Bruno, Pierangela; Greco, Gianluigi; Calimeri, FrancescoProviding accurate diagnoses of diseases and maximizing the effectiveness of treatments requires, in general, complex analyses of many clinical, omics, and imaging data. Making a fruitful use of such data is not straightforward, as they need to be properly handled and processed in order to successfully perform medical diagnosis. This is why Artificial Intelligence (AI) is largely employed in the field. Indeed, in recent years, Machine Learning (ML), and in particular Deep Learning (DL), techniques emerged as powerful tools to perform specific disease detection and classification, thus providing significant support to clinical decisions. They gained a special attention in the scientific community, especially thanks to their ability in analyzing huge amounts of data, recognizing patterns, and discovering non-trivial functional relationships between input and output. However, such approaches suffer, in general, from the lack of proper means for interpreting the choices made by the learned models, especially in the case of DL ones. This work is based on both a theoretical and methodological study of AI techniques suitable for the biomedical domain; furthermore, we put a specific focus on the practical impact on the application and adaptation of such techniques to relevant domain. In this work, ML and DL approaches have been studied and proper methods have been developed to support (i) medical imaging diagnostic and computer-assisted surgery via detection, segmentation and classification of vessels and surgical tools in intra-operative images and videos (e.g., cine-angiography), and (ii) data-driven disease classification and prognosis prediction, through a combination of data reduction, data visualization and classification of high-dimensional clinical and omics data, to detect hidden structural properties useful to investigate the progression of the disease. In particular, we focus on defining a novel approach for automated assessment of pathological conditions, identifying latent relationships in different domains and supporting healthcare providers in finding the most appropriate preventive interventions and therapeutic strategies. Furthermore, we propose a study about the analysis of the internal processes performed by the artificial networks during classification tasks, with the aim to provide a AI-based model explainability. This manuscript is presented in four parts, each focusing on a special aspect of DL techniques and offering different examples of their application in the biomedical domain. In the first part we introduce clinical and omics data along with the popular processing methods to improve the analyses; we also provide an overview of the main DL techniques and approaches aimed at performing disease prediction and prevention and at identifying bio-markers via biomedical data and images. In the second part we describe how we applied DL techniques to perform the segmentation of vessels in the ilio-femoral images. Furthermore, we propose a combination of multi-instance segmentation network and optical flow to solve the multiinstance segmentation and detection tasks in endoscopic images. In the third part a combination of data reduction and data visualization techniques is proposed for the reduction of clinical and omics data and their visualization into images, with the aim of performing DL-based classification. Furthermore, we present a ML-based approach to develop a risk model for class prediction from high-dimensional gene expression data, for the purpose of identifying a subset of genes that may influence the survival rate of specific patients. Eventually, in the fourth part we provide a study on the behaviour of AI-based systems during classification tasks, such as image-based disease classification, which is a widely studied topic in the recent years; more in detail, we show how DL-based systems can be studied with the aim of identifying the most relevant elements involved in the training processes and validating the network’s decisions, and possibly the clinical treatment and recommendation.Item Design and implementation of a modern ASP grounder(2018-01-19) Zangari, Jessica; Leone, Nicola; Calimeri, Francesco; Perri, SimonaAnswer Set Programming (ASP) is a declarative programming paradigm proposed in the area of non-monotonic reasoning and logic programming in the late '80 and early '90. Thanks to its expressivity and capability of dealing with incomplete knowledge, ASP that became widely used in AI and recognized as a powerful tool for Knowledge Representation and Reasoning (KRR). On the other hand, its high expressivity comes at the price of a high computational cost, thus requiring reliable and high-performance implementations. Throughout the years, a signi cant e ort has been spent in order to de ne techniques for an e cient computation of its semantics. In turn, the availability of e cient ASP systems made ASP a powerful tool for developing advanced applications in many research areas as well as in industrial contexts. Furthermore, a signi cant amount of work has been carried out in order to extend the basic language and ease knowledge representation tasks with ASP, and recently a standard input language, namely ASP-Core-2, has been de ned, also with the aim of fostering interoperability among ASP systems. Although di erent approaches for the evaluation of ASP logic programs have been proposed, the canonical approach, which is adopted in mainstream ASP systems, mimics the de nition of answer set semantics by relying on a grounding module (grounder), that generates a propositional theory semantically equivalent to the input program, coupled with a subsequent module (solver ) that applies propositional techniques for generating its answer sets. The former phase, called grounding or instantiation, plays a key role for the successful deployment in real-world contexts, as in general the produced ground program is potentially of exponential size with respect to the input program, and therefore the subsequent solving step, in the worst case, takes exponential time in the size of the input. To mitigate these issues, modern grounders employ smart procedures to obtain ground programs signi cantly smaller than the theoretical instantiation, in general. This thesis focuses on the ex-novo design and implementation of a new modern and e cient ASP instantiator. To this end, we study a series of techniques geared towards the optimization of the grounding process, questioning the techniques employed by modern grounders with the aim of improving them and introducing further optimization strategies, which lend themselves to the integration into a generic grounder module of a traditional ASP system following a ground & solve approach. In particular, we herein present the novel system I-DLV that incorporates all these techniques leveraging on their synergy to perform an e cient instantiation. The system features full support to ASP-Core-2 standard language, advanced exibility and customizability mechanisms, and is endowed with extensible design that eases the incorporation of language upi dates and optimization techniques. Moreover, its usage is twofold: besides being a stand-alone grounder, it is also a full- edged deductive database engine. In addition, along with the solver wasp it has been integrated in the new version of the widespread ASP system DLV recently released.Item Tools and Techniques for Easing the Application of Answer Set Programming(2018-01-19) Fuscà, Davide; Leone, Nicola; Calimeri, Francesco; Perri, SimonaAnswer Set Programming (ASP) is a well-established declarative problem solving paradigm; it features high expressiveness and the ability to deal with incomplete knowledge, so it became widely used in AI and it is now recognized as a powerful tool for knowledge representation and reasoning (KRR). Thanks to the expressive language and the availability of diverse robust systems, Answer Set Programming has recently gained popularity and has been applied fruitfully to a wide range of domains. This made clear the need for proper tools and interoperability mechanisms that ease the development of ASP-based applications. Also, the spreading of ASP from a strictly theoretical ambit to more practical aspects requires additional features for easing the interoperability and integration with other software; furthermore, improving the performance of actual ASP system is crucial for allowing the use of the potential of ASP in new practical contexts. The contribution of this thesis aims at addressing such challenges; we introduce new tools and techniques for easing the application of ASP. In particular, we present EMBASP: a framework for the integration of ASP in external systems for general applications to different platforms and ASP reasoners. The framework features explicit mechanisms for two-way translations between strings recognisable by ASP solvers and objects in the programming language. Furthermore, we define proper means for handling external computations in ASP programs, and implement a proper framework for explicit calls to Python scripts via external atoms into the ASP grounder I-DLV. We also define and implement, into the same system, an additional framework for creating ad-hoc directives for interoperability and make use of it for providing some ready-made ones for the connection with relational and graph databases. Eventually, we work at improving the ASP computation, and present two new ASP systems: DLV2 and I-DLV+MS. DLV2 updates DLV with modern evaluation techniques, combining I-DLV with the solver wasp, while I-DLV+MS is a new ASP system that integrates I-DLV, with an automatic solver selector for inductively choose the best solver, depending on some inherent features of the instantiation produced by I-DLV.