1,721,058 research outputs found
Explainable deep learning for medical image processing: computer-aided diagnosis and robot-assisted surgery.
Full text linkThe recent advancements in the surging field of Deep Learning (DL) have revolutionized every sphere of life, and the healthcare domain is no exception. The enormous success of DL models, particularly with image data, has led to the development of several computeraided diagnosis and clinical support systems. These intelligent imaging systems can help physicians in numerous medical tasks including classification and staging of the various diseases, image-guided surgical procedures, and many more. Additionally, the proliferation of medical datasets has further facilitated the applications of DL techniques in healthcare realm. Moreover, all the perks DL offers are remarkable, however, DL architectures are typically blackbox, i.e. they hide the decision making mechanism, therefore, interpreting how the model arrived at a particular decision is hidden. Additionally, Convolutional Neural Networks (CNNs), which are most widely used DL techniques, are prone to adversarial examples, where small, imperceptible perturbations to the input data can cause the model to make incorrect predictions. These facts question the applicability of DL in healthcare sector where explainability holds paramount significance to build a trust on surging field of machine learning. The concept of eXplainable Artificial Intelligence (XAI) brings forward the possibility of explaining the results of DL models and reveals how the models produce results. These techniques aim to improve the transparency and interpretability of AI models, which can enhance trust in their results and facilitate their adoption in clinical practice. XAI approaches have the potential to advance the understanding of complex medical image analysis tasks and improve the reliability of AI-based diagnosis and treatment planning. The story does not end here, the XAI methods in the context of medical imaging generally produce saliency maps and compute feature importance to explain the results of DL models. The sensitive nature of healthcare industry, because of having the direct correlation with human life, questions the authenticity of XAI outcomes, and demands a qualitative and quantitative measure to evaluate these evaluation methods. Furthermore, heatmap visualizations alone are often insufficient for achieving transparency and interpretability of DL models in medical imaging to foster the AI and biomedical synergy. Inspired by the latest trends and contributions in light of the aforementioned concerns, this thesis designs, develops, and validates an interpretable and transparent intelligent clinical decision support system based on traditional machine and DL architectures, whose outcomes can be qualitatively and quantitatively explained with XAI methods. The thesis also comprises a segmentation and detection pipeline for image-driven surgical applications. These novel intelligent systems aims to assist the physicians and clinicians in image-guided diagnostic and treatment systems. The developed interpretable diagnostic frameworks offer wide range of applications and can be extended to several clinical scenarios. Concerning the XAI, transparency and interpretability of CNN architectures are achieved through two families of XAI methods, i.e. perceptive and mathematical XAI. Furthermore, within each of these XAI families, two explanation frameworks are employed. These methods facilitated to investigate the reliability of features and learning process, to critically analyse various CNN architectures and XAI methods, and to compare the outcomes of both XAI pipelines. To further highlight the applications of DL in the image-guided surgical domain, a case study has been performed on image-guided surgical procedures and interventions. The case study also encompasses a detailed investigative study of public datasets and presents the legal and ethical issues of DL-driven image-guided surgery. The study additionally underlines the risks and limitations towards the autonomous systems and provides the future perspective. Finally, the second case study investigates the qualitative and quantitative evaluation of the XAI techniques in regards to the medical images. The case study also sheds light on the evaluation measures, metrics for XAI, quality of explanation, types of explanation, and few more. The clinical efficacy of the developed solutions is evaluated through comparison with existing state-of-the-art methods, and is further validated through consultation with physicians where feasible. The datasets incorporated during the study are either obtained from the online open source platforms or collected from local health institutions
Le nuove regole di Basilea2: prime valutazioni di impatto sul rapporto banca/impresa in Italia
No full textLa rischiosità delle imprese italiane: indicazioni dall’applicazione dei modelli di rating interno
No full textL'analisi effettuata applicando modelli di previsione delle insolvenze su un campione di oltre 80 mila imprese per il triennio 2003-2005 ha evidenziato una significativa differenziazione nell'evoluzione delle condizioni di rischiosità delle imprese. Per il segmento Corporate le stime segnalano una riduzione della rischiosità per le imprese di media dimensione e per quelle retail un innalzamento delle probabilità di default.Tali indicazioni sono confermate anche dalle analisi di migrazione delle imprese tra le diverse classi di rating. L'analisi sulle determinanti nelle variazioni dei rating segnala che per le Sme e per le Corporate gli effetti negativi del deterioramento della redditività operativa sono stati più che compensati dal miglioramento degli indici relativi alla struttura finanziaria del servizio del debito
L’impatto sul sistema bancario dell’avvio di Basilea2: un’analisi empirica
No full textIl 1° gennaio del 2008 è entrata pienamente in vigore la nuova disciplina prudenziale sul capitale delle banche (nota come Basilea 2) che prevede – tra l’altro – la possibilità per le banche di calcolare il requisito patrimoniale di vigilanza a fronte del rischio di credito sulla base di sistemi di rating sviluppati internamente e validati dalle autorità di vigilanza. Il presente lavoro stima a tutto il 2007 le condizioni di rischiosità di un campione di imprese italiane in base alle indicazioni fornite da modelli statistici di determinazione del Rating delle singole imprese simili a quelli previsti dal Nuovo Accordo sul Capitale per il calcolo dei requisiti di vigilanza prudenziale. Estendendo ed affinando la metodologia utilizzata in precedenti analisi (Bocchi-Lusignani, 2004, 2006) si perviene alla stima delle probabilità di default in ciascuno dei 5 anni del periodo 2003-2007 e alla quantificazione dell’impatto sui requisiti di capitale richiesti alle banche, secondo quanto previsto dall’applicazione dell’approccio FIRB, confrontandolo con quello previsto dall’approccio standardizzato. L’applicazione di modelli di previsione delle insolvenze ad un campione di oltre 80 mila imprese per il quinquennio 2003-2007, ha confermato una significativa differenziazione nell’evoluzione delle condizioni di rischiosità delle imprese. Per le imprese di maggiore dimensione (segmento Corporate) le stime segnalano una riduzione della rischiosità mentre, al contrario, una sostanziale stabilità per le imprese di media dimensione (SME) ed un innalzamento nel valore delle probabilità di default per quelle di dimensione più piccola (Retail), soprattutto nel periodo più recente. Per il segmento SME l’analisi sulle determinanti nelle variazioni dei rating segnala che gli effetti negativi del deterioramento della redditività operativa sono stati più che compensati dal miglioramento degli indici relativi alla struttura finanziaria e al servizio del debito. Al contrario sul deterioramento della rischiosità delle imprese di più piccola dimensione hanno pesato in misura più significativa la riduzione di redditività ed il basso livello di autofinanziamento solo parzialmente compensata dalla riduzione degli oneri finanziari derivante dal livello particolarmente basso dei tassi di interesse nel periodo. Le stime di impatto sul requisito patrimoniale ottenute incorporando nelle funzioni di ponderazione regolamentari i valori delle probabilità di default delle imprese stimate per il campione confermano, in tutti i periodi di osservazione, un valore del requisito patrimoniale normalizzato (incorporando anche l’ammontare delle perdite attese) richiesto alle banche inferiore a quello attuale dell’8% per le imprese SME Retail e per le imprese Corporate, mentre per il segmento delle imprese SME Corporate (fra i 5 e i 50 milioni di euro di fatturato) il requisito patrimoniale normalizzato risulta prossimo ai valori dell’8%. L’inferenza sul totale delle imprese del sistema, pur con tutte le cautele interpretative dovute a una non piena rappresentatività del campione analizzato, conferma valori medi del requisito patrimoniale per il rischio di credito inferiori a quelli del requisito standardizzato, anche se il divario per alcuni comparti potrebbe non essere sufficiente a coprire le necessità di capitale per fronteggiare i rischi operativi e quindi richiedere un livello del requisito superiore a quello corrente
A multipurpose user-friendly tool for voice analysis : application to pathological adult voices
No full textAssessing voice quality objectively is of great relevance to clinicians, both for quantifying surgical or pharmacological effectiveness and for detecting and classifying voice pathology. A large number of objective indexes have been proposed in literature and implemented in commercially available software tools. However, clinicians commonly resort to a small subset of these indexes since they may be difficult to set up or understand. This paper presents a new user-friendly voice analysis tool named BioVoice. At present, BioVoice allows for the evaluation of few but important indexes, devoting great effort to their robust and automatic evaluation, although extensions are foreseeable. Specifically, fundamental frequency, along with irregularity (jitter, relative average perturbation), noise, and formant frequencies, is tracked on voiced parts of the signal only. Mean and standard deviation values are also calculated and displayed. This high-resolution estimation procedure is further strengthened by an adaptive estimation of the optimal length of signal frames for analysis, linked to varying signal characteristics. Moreover, BioVoice allows automatic analysis of any kind of voice signal as far as F0 range and sampling frequency are concerned, with no manual setting required. This new tool is thus feasible for use by non-experts from different scientific fields, thanks to its simple interface. Here, the proposed approach is applied to patients who underwent micro-laryngoscopic direct exeresis to remove cysts and polyps. Pre- and post-surgical indexes were estimated using BioVoice and then compared with the output of one of the most common commercial software tools to both assess voice quality recovery and to evaluate the new method's capabilities
From architectural to behavioural specification of services
No full textMany efforts are currently devoted to provide software developers with methods and techniques that can endow service-oriented computing with systematic and accountable engineering practices. To this purpose, a number of languages and calculi have been proposed within the Sensoria project that address different levels of abstraction of the software engineering process. Here, we report on two such languages and the way they can be formally related within an integrated approach that can lead to verifiable development of service components from more abstract architectural models of business activities
Going Beyond Counting First Authors in Author Co-citation Analysis
Full text linkThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
- …
