IRIS UniSR (’Università Vita-Salute San Raffaele)
Not a member yet
54666 research outputs found
Sort by
A network-based approach reveals higher plasticity levels in bipolar than major depressive disorder
No full textBackground: Bipolar disorder (BD) and major depressive disorder (MDD) are often misdiagnosed due to overlapping symptomatology, leading to inappropriate treatments and poor outcomes. Thus, novel approaches for effective early differential diagnosis are warranted. Plasticity – the ability to modify brain functioning and mental state – is increasingly recognized as a crucial process in psychiatry, as it underlies transitions from psychopathology to wellbeing. Accordingly, enhanced plasticity is reported to increase mental state transitions. Since BD patients experience transitioning between depression and mania, we hypothesized this disorder to be characterized by higher plasticity levels than MDD. Methods: We leveraged a recent network-based approach to measure plasticity, operationalizing it as the inverse of the connectivity strength among network symptoms. We analyzed 211 BD and 136 MDD patients undergoing a major depressive episode. Symptoms were employed to generate unregularized networks via Gaussian Graphical Model, with Spearman's Rank correlation estimating the input matrix. Connectivity strength was compared between disorders using the Network Comparison Test function. Results: BD patients exhibit higher plasticity levels than MDD patients, as demonstrated by weaker symptom connectivity, despite both groups having comparable depressive symptom severity (p = 0.008). Within the BD group, type I patients showed also higher plasticity compared to type II (p = 0.04), further corroborating the association between high plasticity and mood instability. Conclusions: Our findings support the hypothesis that plasticity levels differ across psychopathological conditions based on their likelihood of mental transitions. Consequently, measuring of plasticity through symptom network dynamics holds promise for the early differential diagnosis of BD and MDD
Niccolò Cusano e le metamorfosi del neoplatonismo. La rinascita di Proclo e Dionigi
No full textCusano è il pensatore rinascimentale influenzato in modo più sistematico da Proclo e Dionigi Areopagita, che costituiscono i cardini attorno ai quali costruisce la sua interpretazione del platonismo. Il neoplatonismo di Proclo e Dionigi è abitato congiuntamente da due tipi di metafisiche: la metafisica gerarchica e la metafisica dell’ineffabile. La prima comporta l’idea che la realtà sia una struttura composta da gradi subordinati, che vanno dal Principio divino al nostro mondo, passando anche per livelli intermedi. La metafisica dell’Ineffabile, invece, si riferisce alla natura stessa del Principio divino, da cui scaturisce l’intera gerarchia e che è però al contempo al di là della struttura gerarchizzante stessa. Considerato nella sua natura più pura, il Principio divino – ineffabile in quanto al di sopra delle leggi della logica e, in ultimo, dello stesso principio di non contraddizione – è caratterizzato da assoluta semplicità e arelazionalità, al punto che, per Proclo, è superiore sia all’Uno che al Bene e per Dionigi persino alla stessa dimensione trinitaria. La ricezione in Cusano di Proclo e Dionigi va vista nei termini di uno sbilanciamento – e, anzi, di una rottura – della metafisica gerarchica in favore della metafisica dell’Ineffabile. Ciò che Cusano fa, infatti, è sottolineare che, essendo il Principio divino al di sopra delle leggi della logica e della stessa scienza, e costituendo la natura intima di ogni essere, allora la stessa strutturazione gerarchica della realtà, che aveva per il neoplatonismo antico una validità metafisica, ha piuttosto una validità epistemologico-mentale. Tutta la realtà, nella sua intima natura, in quanto abitata dal Principio divino, è ineffabile, e la strutturazione gerarchica della scienza ha una natura congetturale – cioè, pur avendo essa a che fare con la verità, della quale partecipa, non c’è più coincidenza tra pensiero e realtà. Nelle metamorfosi del neoplatonismo in Cusano, attraverso lo sviluppo della metafisica dell’Ineffabile, vediamo quindi: 1) come questa tradizione anticipi lo sviluppo del problema della conoscenza che diventerà poi centrale nella tradizione filosofica europea. Inoltre, 2) la rottura della metafisica gerarchica comporta anche un avvicinamento tra Dio e il mondo, che, pur non diventando panteismo, batte l’accento in modo inaudito sull’immanenza del Principio divino in ogni ente. Infine, 3) la congetturalizzazione della metafisica ha conseguenze per Cusano anche a livello di analisi storica del rapporto tra culture diverse. Poiché ogni tradizione metafisica, così come ogni tradizione religiosa, ha una natura congetturale – anche se ha sempre a che fare con l’unico Principio divino meta-congetturale – ne consegue che nessuna delle diverse tradizioni, nemmeno quella cristiana, può pretendere di essere il possessore della verità. Pertanto, per natura, ogni tradizione deve intrinsecamente dialogare con le altre.Cusanus is the Renaissance thinker most systematically influenced by Proclus and Dionysius the Areopagite, who form the cornerstones around which he builds his interpretation of Platonism. Proclus’ and Dionysius’ Neoplatonism is jointly inhabited by two types of metaphysics: hierarchical metaphysics and the metaphysics of the ineffable. The former entails the idea that reality is a structure composed of subordinate degrees, ranging from the divine Principle to our world, passing also through intermediate degrees. The metaphysics of the Ineffable, on the other hand, refers to the nature of the divine Principle itself, from which the entire hierarchy stems and yet is at the same time beyond the hierarchising structure itself. Considered in its purest nature, the divine Principle – ineffable insofar as it is above the laws of logic and, ultimately, the principle of non- contradiction itself – is characterised by absolute simplicity and areationality, to the point that, for Proclus, it is superior to both the One and the Good, and for Dionysius even to the Trinitarian dimension itself. The reception in Cusanus of Proclus and Dionysius should be seen in terms of an unbalancing – and indeed a rupture – of hierarchical metaphysics in favour of the metaphysics of the Ineffable. What Cusanus does, in fact, is to emphasise that since the divine Principle is above the laws of logic and science itself, and constitutes the intimate nature of all being, then the same hierarchical structuring of reality, which had for ancient Neoplatonism a metaphysical validity, has rather an epistemological-mental validity. All reality, in its intimate nature, as inhabited by the divine Principle, is ineffable, and the hierarchical structuring of science has a conjectural nature – that is, although it has to do with truth, in which it participates, there is no longer any coincidence between thought and reality. In the metamorphoses of Neoplatonism in Cusanus, through the development of the metaphysics of the Ineffable, we thus see: 1) how this tradition anticipates the development of the problem of knowledge that will later become central to the European philosophical tradition. Furthermore, 2) the break with hierarchical metaphysics also entails a rapprochement between God and the world, which, while not becoming pantheism, places unprecedented emphasis on the immanence of the divine Principle in every being. Finally, 3) the conjecturalisation of metaphysics also has consequences for Cusanus at the level of historical analysis of the relationship between different cultures. Since every metaphysical tradition, like every religious tradition, is conjectural in nature – even if it always has to do with the one meta-conjectural divine Principle – it follows that none of the different traditions, not even the Christian tradition, can claim to be the possessor of truth. Therefore, by nature, each tradition must inherently dialogue with the others
Nonsuicidal Self-injury in the Perspective of HiTOP Spectra
No full textIntroduction: Nonsuicidal self-injury (NSSI) represents a relevant public health concern, with lifetime prevalence being high in community samples. The present study aimed to examine the latent associations between the Hierarchical Taxonomy of Psychopathology (HiTOP) superspectra and NSSI frequency and motivation. Methods: A sample of 547 community-dwelling adult participants was administered measures of NSSI, NSSI functions, and psychopathology. Results: The multiple indicators multiple causes model evidenced a significant and nontrivial contribution of the Functional Assessment of Self-Mutilation automatic function latent dimension in predicting the frequency of NSSI. Structural equation modeling analyses showed that the overall frequency of NSSI episodes was uniquely, significantly and positively predicted by the HiTOP Externalizing latent dimension scores. Notably, all FASM motivation factors yielded significant and nontrivial relationships with HiTOP Externalizing, Psychosis, and Emotion Dysfunction latent variables in SEM analyses. Conclusions: These findings may prove useful in extending our knowledge of transdiagnostic psychopathology dimensions and their implications for NSSI
Clinical integration of brain and cord MRI features improves differential diagnosis of multiple sclerosis
No full textObjective: To explore the role of brain and spinal cord MRI features in differentiating patients with suspected central nervous system (CNS) inflammatory diseases. Material and methods: Prospective data from 125 patients undergoing diagnostic evaluation, including 1.5 T brain and spinal cord MRI scans from February 2021 and March 2024 were analyzed. The cohort comprised 91 patients with multiple sclerosis (MS), 15 with other inflammatory neurological diseases (OIND), and 19 with non-inflammatory neurological diseases (NIND). Brain and spinal cord lesion topographies and morphological features were evaluated to identify MRI features discriminating MS from OIND and NIND. Results: Random forest analysis identified key MRI features supporting MS diagnosis over OIND: absence of longitudinally extensive transverse myelitis (relative importance [RI] = 100%), presence of ≥ 1 Dawson’s finger (RI = 55.3%), ≥ 1 cortical lesion (RI = 42.6%), and ≥ 1 brain T2-hyperintense white matter (WM) lesion (RI = 36.4%). After excluding the presence of ≥ 1 brain T2-hyperintense WM lesion, fulfilling ≥ 2 of the 3 selected criteria distinguished MS from OIND patients with a sensitivity of 0.59 and a specificity of 0.80. For distinguishing MS from NIND, relevant MRI features included ≥ 1 T2-hyperintense spinal cord lesion (RI = 100.0%), ≥ 1 Dawson’s finger (RI = 84.3%), ≥ 1 cortical lesion (RI = 61.4%), ≥ 1 cerebellar peduncle lesion (RI = 52.2%) and ≥ 3 central vein sign-positive lesions (RI = 27.8%). Fulfilling ≥ 2 of the 5 selected criteria identified MS patients with a sensitivity of 0.64 and a specificity of 0.84. Conclusion: Integrating specific MRI features in the diagnostic work-up of patients with suspected CNS inflammatory disease improves differentiation between MS, OIND, and NIND, reducing the risk of misdiagnosis
Model of End-Stage Liver Disease–alpha-fetoprotein–tumor burden (MELD-AFP-TBS) score to stratify prognosis after liver resection for hepatocellular carcinoma
No full textIntroduction: Morphologic criteria, such as the Barcelona Clinic Liver Cancer staging system often fail to accurately predict long-term survival among patients undergoing liver resection for hepatocellular carcinoma. We sought to develop a continuous risk score that incorporates established markers of tumor biology and liver function to improve the prediction of overall survival. Methods: Data from a multi-institutional database were used to identify patients who underwent curative-intent hepatectomy for hepatocellular carcinoma. A predictive score for overall survival was developed using weighted beta-coefficients from a multivariable Cox regression model. Results: Among 850 patients, 595 (70.0%) were assigned to the training cohort, and 255 (30.0%) to the test cohort. In the training cohort, multivariable analysis identified the Model of End-Stage Liver Disease (hazard ratio, 1.04; 95% confidence interval, 1.01-1.07), log-transformed alpha-fetoprotein (hazard ratio, 1.07; 95% confidence interval, 1.02-1.13), and tumor burden score (hazard ratio, 1.07; 95% confidence interval, 1.03-1.11) as independent predictors of worse overall survival. The Model of End-Stage Liver Disease-alpha-fetoprotein-tumor burden score, based on the Cox model, stratified patients into low-risk (n = 466, 78.3%) with a 5-year OS of 70.5% and high-risk (n = 129, 21.7%) with a 5-year OS of 47.0% (P < .001). In the test cohort, the Model of End-Stage Liver Disease-alpha-fetoprotein-tumor burden score demonstrated superior discriminative accuracy (C-index: 0.72, time-dependent area under the curve 1-year: 0.80, 3-year 0.76, 5-year 0.70) compared with the Barcelona Clinic Liver Cancer staging system (C-index: 0.53, time-dependent area under the curve 1-year: 0.61, 3-year 0.55, 5-year 0.56). An online tool was made accessible at https://jk-osu.shinyapps.io/MELD_AFP_TBS/. Conclusion: The Model of End-Stage Liver Disease-alpha-fetoprotein-tumor burden score provides a novel, accurate tool for prognostic stratification of patients with hepatocellular carcinoma, identifying high-risk patients who may benefit from alternative treatments to improve outcomes
Radiomics and Artificial Intelligence in medical imaging for precision medicine
No full textBackground and Purpose: Advances in imaging technologies and artificial intelligence (AI) algorithms have driven progress in AI-based image analysis and radiomics. These emerging fields focus on extracting quantitative data from medical images, offering deeper insights into disease characterization and prognosis prediction. This PhD project aims to advance radiomics, addressing key clinical questions in oncological and non-oncological settings to support tailored patient care.
Materials and Methods. To achieve the overall aim, different methodological aspects have been systematically explored, encompassing quantitative imaging parameters (AIM 1 and 2), handcrafted radiomics-HCR (AIM 3 and AIM 4), and deep radiomics (AIM 5). This was applied in different clinical settings, including oncology and inflammation. In AIM 1, the preoperative role of [68Ga]Ga-DOTATOC PET conventional parameters in predicting DAXX/ATRX loss of expression in 72 patients with pancreatic neuroendocrine tumors (PanNET) was investigated. In AIM 2, the role of machine learning (ML)-based classification using conventional [18F]FDG PET parameters was evaluated for the preoperative prediction of tumor aggressiveness in a cohort of 123 endometrial cancer (EC) patients. Subsequently, the role of HCR was explored, with AIM 3 aiming at improving lymph nodes (LN) detection in 72 PanNETs patients, and AIM 4 targeting the prediction of p53 hyperexpression in 121 EC patients. In all these oncological settings, post-surgical histology was used as the reference standard. Finally, in AIM 5, the role of [18F]FDG PET-based convolutional neural networks (CNNs) in predicting Takayasu (TAK) arteritis flare-ups within 12 months from the scan was investigated.
Results. AIM 1: PET-derived somatostatin-receptor density (SRD) was predictive for DAXX loss of expression before surgery (AUC=79.01%; cut-off=46.96), with an increase of one unit of SRD associated with approximately a 4% increase in the odds of DAXX LoE (p=0.002). AIM 2: Best accuracies for individual PET parameters and ML were: 61% (TLG) and 87% (ML) for myometrial invasion; 71% (SUVmax) and 79% (ML) for risk groups; 72% (TLG) and 83% (ML) for LN; 45% (SUVmax; SUVmean) and 73% (ML) for p53 expression. Overall, ML models consistently improved all the investigated predictions. AIM 3: [68Ga]Ga-DOTATOC PET qualitative examination of LN involvement provided bACC= 60% (SN= 24%, SP= 95%); the best-performing radiomic model provided a bACC= 70%, allowing to significantly increase detection SN (SN=77%, SP= 61%). AIM 4: Prediction of p53 hyperexpression achieved bACC=0.78 (SN=0.83, SP=0.64) with a stratified train/test split and bACC=0.60 (SN=SP=0.60) using Monte Carlo cross-validation. None of the tested oversampling, selection, or ML approaches improved results further. AIM 5: The proposed liver-based standardization approach for PET harmonization was the only method enabling complete comparison among scans from different tomographs. For the prediction of TAK flare-ups, the best CNN model achieved 82% accuracy, 85% precision of class 0, 62% precision of class 1, 38% recall, 94% specificity and 70% nMCC.
Conclusions. This PhD project assessed the potential and challenges of quantitative image analysis, radiomics and AI in advancing personalized medicine. While demonstrating their value in improving disease prediction and characterization in both oncological (PanNETs, EC) and non-oncological (TAK) settings, it also highlighted critical issues, including the need for robust methodologies and the lack of standardization across many methodological approaches.Obiettivi: I progressi nelle tecnologie di imaging e negli algoritmi di intelligenza artificiale (AI) hanno favorito lo sviluppo dell'analisi delle immagini mediche e della radiomica, discipline che sfruttano dati quantitativi per caratterizzare malattie e prevederne la prognosi. Questo progetto di dottorato mira a promuovere l’evoluzione della radiomica, affrontando quesiti clinici cruciali in ambiti oncologici e non, per supportare cure personalizzate.
Metodi: Diversi aspetti metodologici sono stati esplorati, includendo parametri di imaging quantitativi (OBIETTIVI 1 e 2), radiomica hand-crafted (HCR) (OBIETTIVI 3 e 4) e deep radiomics (OBIETTIVO 5). Questi approcci sono stati applicati a diversi contesti clinici, tra cui oncologia e infiammazione. Nell’OBIETTIVO 1 è stato analizzato il ruolo preoperatorio dei parametri convenzionali PET [68Ga]Ga-DOTATOC nella previsione della perdita di espressione di DAXX/ATRX in 72 pz con tumori neuroendocrini pancreatici (PanNET). Nell’OBIETTIVO 2 è stato valutato il ruolo del machine learning (ML) e parametri convenzionali PET [18F]FDG per la predizione preoperatoria dell’aggressività tumorale in una coorte di 123 pz con carcinoma endometriale (EC). Il ruolo della HCR è stato esplorato con l’OBIETTIVO 3, focalizzato sul miglioramento del rilevamento linfonodale (LN) in 72 pz PanNET, e l’OBIETTIVO 4, dedicato alla predizione dell’iper-espressione di p53 in 121 pz EC. In tutti questi contesti, l’istologia post-chirurgica è stata usata come standard di riferimento. Nell’OBIETTIVO 5 è stato indagato il ruolo delle reti neurali convoluzionali (CNN) basate su PET [18F]FDG nella previsione delle riacutizzazioni dell’arterite di Takayasu (TAK) entro 12 mesi dalla PET.
Risultati: OB 1: Il parametro SRD estratto dalla PET è risultato predittivo della perdita di espressione di DAXX (DAXX LoE): AUC=79,01% (cut-off=46,96), con un aumento di una unità di SRD associato a un aumento di circa 4% della probabilità di DAXX LoE (p=0.002). OB 2: Le accuratezze di predizione ottenute per i singoli parametri PET e i modelli ML sono state: 61% (TLG) e 87% (ML) per l’invasione miometriale; 71% (SUVmax) e 79% (ML) per i gruppi di rischio; 72% (TLG) e 83% (ML) per i LN; 45% (SUVmax; SUVmean) e 73% (ML) per l’espressione di p53. In tutte le analisi, i modelli ML hanno apportato miglioramenti nella predizione degli outcome. OB 3: L’esame qualitativo della PET [68Ga]Ga-DOTATOC per il coinvolgimento dei LN ha fornito una bACC di 60% (SN=24%, SP=95%); il miglior modello radiomico ha raggiunto bACC=70%, aumentando significativamente la sensibilità (SN=77%, SP=61%). OB 4: La predizione dell’iper-espressione di p53 ha ottenuto bACC=0.78 (SN=0.83, SP=0.64) con uno split stratificato train/test, e bACC=0.60 (SN=SP=0.60) utilizzando Monte Carlo Cross Validation. Nessun metodo di oversampling, selezione, o modelli ML ha migliorato ulteriormente i risultati. OB 5: L’approccio di standardizzazione basato sul fegato proposto per l’armonizzazione delle PET è stato l’unico metodo a consentire un confronto completo tra gli scan provenienti da diversi tomografi. Per la previsione delle riacutizzazioni di TAK, il miglior modello CNN ha raggiunto un’accuratezza dell’82%, precisione per la classe 0 dell’85%, per la classe 1 del 62%, recall 38%, specificità del 94% e nMCC del 70%.
Conclusioni: Questo progetto ha valutato il potenziale e le sfide dell’analisi quantitativa delle immagini, della radiomica e dell’AI nel supportare la medicina personalizzata. Pur dimostrandone il valore nel migliorare la previsione e la caratterizzazione delle malattie sia in contesti oncologici (PanNET, EC) sia non oncologici (TAK), ha anche evidenziato problematiche critiche, tra cui la necessità di metodologie robuste e la mancanza di standardizzazione
From semantic concreteness to concretism in schizophrenia: An automated linguistic analysis of speech produced in figurative language interpretation
No full textLack of abstract thinking, known as concretism, is a well-known psychopathological feature of schizophrenia, reflecting the tendency to adhere to concrete aspects of stimuli and figurative language comprehension difficulties. Inspired by the similarity between ‘concretism’ as defined in psychopathology and ‘concreteness’ as defined in linguistics, namely a semantic dimension linked to perceptual experience, we tested the novel hypothesis that impairment in deriving figurative meanings is related to impairment at the semantic level, involving concreteness. We analysed speech samples from 63 individuals with schizophrenia and 47 controls, who were asked to verbalise the meaning of idioms, metaphors, and proverbs. By automatically extracting linguistic features from speech, we observed that answers in the schizophrenia group exhibited higher word concreteness and the related measure of word imageability, especially in proverbs, while not differing from controls’ ones in lexical richness and speech-time composition. Concreteness in verbalisations produced by individuals with schizophrenia negatively predicted their ability to understand proverbs and their global pragmatic and cognitive profile. This study supports the idea that concretism is rooted in semantics, linking the tendency to concrete figurative interpretations and a bias towards concrete words. In this view, impairment in figurative language understanding can be seen as a difficulty in abstracting away from perceptual-related properties associated with linguistic inputs, in the broader context of multisensory integration disruption. The study discloses new areas of interest for the automated analysis of speech in psychosis, pointing to the importance of considering concreteness for better characterising linguistic profiles and identifying clinically relevant linguistic dimensions
Persistence of CXCR4-tropic virus in people living with four-class drug-resistant HIV and its clinical impact in the modern antiretroviral era
No full textBackground CXCR4-tropic HIV seems to be associated with more clinical events than CCR5-tropic virus. Objectives This study aims to describe the effect of the persistence of CXCR4-tropic virus on the occurrence of clinical events in people with four-class drug-resistant HIV. Methods This is a retrospective study on people with four-class drug-resistant HIV from the PRESTIGIO Registry, with at least two HIV-tropism determinations during follow-up. Follow-up accrued from the date of the first four-class drug resistance evidence (baseline) until death, loss to follow-up or freezing date (31 December 2023). Univariable Poisson regression was used to estimate and compare incidence rates of clinical events. Predictors of clinical events were assessed by multivariable Poisson regression. Results A total of 144 people with four-class drug-resistant HIV [47 (33%) with persistent CXCR4-tropism, 39 (27%) with persistent CCR5-tropism and 58 (40%) with a tropism switch during follow-up] were included with a median follow-up of 7.80 years (IQR = 5.80-10.6). Overall, 117 (81.3%) 4DR-PLWH experienced at least one clinical event during follow-up [incidence rate = 32.5 (95% CI = 29.3-35.9)]. The persistence of CXCR4-tropic virus was associated with an increased risk of HIV-related events among people living with four-class drug-resistant HIV, even in modern ART era. After adjusting for age, sex at birth and CD4+/CD8+ at baseline, standardized viremia copy-years [adjusted-incidence rate ratio = 1.66 (95% CI = 1.24-2.26), P < 0.001] and persistent CXCR4-tropism [adjusted-incidence rate ratio: 2.01 (95% CI = 1.04-3.91), P = 0.037] were associated with the occurrence of HIV-related events. Conclusions Our findings confirm CXCR4-tropism as a marker of HIV progression also in the four-class drug-resistant population, suggesting the need of further prioritization of viro-immunological control and studies of pathogenic mechanisms in presence of CXCR4-tropic multidrug-resistant viral strains
Tumor-specific major histocompatibility-II expression predicts pathological complete response to atezolizumab combined to chemotherapy in triple-negative breast cancer
No full textAdding immune checkpoint inhibitors to neoadjuvant chemotherapy improves outcomes in early-stage triple-negative breast cancer (TNBC), but a fraction of patients derive benefit. Tumor-specific MHC-II (tsMHC-II) expression has been shown to be a predictive biomarker of pathological complete response (pCR) to neoadjuvant chemo-immunotherapy in early-stage TNBC. We performed biomarker analysis of the phase III NeoTRIP trial where patients were randomized to neoadjuvant carboplatin and nab-paclitaxel±atezolizumab. Imaging mass cytometry was used to assess tsMHC-II expression in tumor samples. TsMHC-II positivity was predefined as ≥5% of tumor cells expressing MHC-II, and at an 80th percentile exploratory cutoff. TsMHC-II positivity was associated with a higher pCR rate in the atezolizumab arm (OR:2.58; P = 0.016), but not in the chemotherapy-only arm (OR:1.37; P = 0.34) and these results were stronger using the exploratory cutoff. TsMHC-II expression is associated with improved response to neoadjuvant chemo-immunotherapy in early TNBC and could represent a clinically useful predictive biomarker for treatment personalization
