Search

Background: non-valvular atrial fibrillation (NVAF) is the most common arrhythmia in clinical practice. Its risk increases with age and the presence of comorbidities. NVAF-related morbility is associated to thromboembolic events and, among patients undergoing to anticoagulant therapy, to haemorrhagic complications. Guidelines suggest stratifying the thrombotic risk with CHA2DS2-VASc score and the haemorrhagic risk with HAS-BLED score before initiating any anticoagulant treatment. These two scores share several items, and it is common to observe patients with both an increased thrombotic and haemorrhagic risk. This area represents a “grey zone” where guidelines are not able to suggest any evidence-based approach. Patients and Methods: we enrolled 519 consecutive, critically-ill subjects affected by permanent NVAF admitted to our Subintensive Medicine department. CHA2DS2-VASc score and HAS-BLED score were calculated at admission. For each patient, we evaluated age, sex, admission diagnosis, comorbidities, thromboembolic and haemorrhagic events. Results: mean age was 75.61(±11.92) years; males represented 50.3% of the sample; patients had a median of 2(0-6) comorbidities. We observed 38 haemorrhagic events and 80 thromboembolic events; HAS-BLED had a median of 2(0-5), CHA2-DS2-VASc score had a median of 3(0-6); CHA2DS2-VASc had an AUC of 0.56;95%CI:0.50-0:63(p=0.06); HAS-BLED had an AUC of 0.53;95%CI:0.44-0:62(p=0.53). Discussion: in this population of elderly patients admitted for critical illnesses, CHA2DS2-VASc and HAS-BLED scores have a limited predictive value for both thromboembolic and haemorrhagic events. Newer approaches based on machine learning and topological data analysis could better discriminate the risk factors of NVAF-related complications and give a more reliable prediction of thrombotic or haemorrhagic events.

Background: The association of systemic sclerosis (SSc) with the exposureto environmental agents is supported by a number of case reports and somecase-control studies. No conclusive results have been reported, but there aresome evidences that exposure to vinyl-chloride-polymers (PVC), silica dustor organic solvents such as benzene (B) and xylene (X) may be implicated.Furthermore a higher prevalence of scleroderma in boroughs in close proximityto a major airport, has been reported, but few data on air pollution exposure andrisk of systemic sclerosis are available. Recently, particulate air pollution hasbeen consistently linked to increased risk of arterial cardiovascular disease.Objectives: We studied relationships between outdoor concentration of B andparticulate with clinical manifestations of SSc based, for the exposure, on theurban residence of patients.Methods: Before patient administration, the questionnaire was validated byDelphi technique (4 rheumatologists, 4 statisticians and 2 common people).A cohort study of 88 SSc patients, filled the validated self administeredquestionnaire (analyzing drug, work and environmental exposure) toinvestigate potential risk exposure before and after the onset of the disease.The average mean concentrations of B (11 monitoring sites) and environmentalparticulate matter with aerodynamic diameter ≤ 10 μm(PM10) (14 sites) werecomputed using data from monitors located throughout the Lazio region, inItaly. In a sample of 33 patients we performed correlations between the concentrationsof PM10 and B with the demographic and clinical characteristics,going back to a prior exposure of 2 years before the onset of Raynaud’s phenomenon(RP).Results: The questionnaire resulted in an agreement of the overall expertsof about 94% (according to11/190 disagreement for comprehension, onlyfew lexical modifications were done to improve the questionnaire afterthe consensus between the experts), with an Inter-observer agreement(measured throughout K Cohen test) of 0.8019(p

Segmentazione di immagini FLAIR assiali basata su analisi topologica dei dati Matteo Rucco (4,6), Antonella Castellano (2,6), Marina Donativi (1,6), Giorgio De Nunzio (1,6), Emanuela Merelli(4), Damir Herman (7), Tanya Petrossian (7), Gabriella Pastore (3,6), Lorenzo Bello (5), Andrea Falini (2) 1. Dip. di Matematica e Fisica “Ennio De Giorgi”, Univ. del Salento, Lecce 2. U.O. Neuroradiologia, Ospedale San Raffaele e Univ. Vita-Salute, Milano 3. Istituto di Ricerche Cliniche Ecomedica, Centro di Radioterapia e IGRT, Empoli 4. Univ. di Camerino, School of Science and Technology, Computer Science Division, Camerino 5. U.O. Neurochirurgia, Ist. Clinico Humanitas, Univ. di Milano, Milano 6. Advanced Data Analysis in Medicine, http://adamgroup.it 7. Ayasdi Inc., Palo Alto - California FINALITA'.La necessitá di sviluppare tecniche innovative per l’analisi di big dataè stata in parte soddisfatta con lo sviluppo di algoritmi che trovano origine nella topologia algebrica e forniscono una rappresentazione compatta del dataset iniziale ed enfatizzano correlazioniche metodi statistici classici offuscherebbero. In questo studio abbiamo applicato l’algoritmo ”Mapper”[1] implementato nel software Iris che fornisce la rappresentazione in forma di un grafo costituito da nodi (cluster).La rappresentazione a grafo è ottenuta proiettando le coordinate dei pixel in uno spazio metrico, i nuovi pixel sono collegati fra loro per similaritá di toni di grigio di immagini FLAIR assiali. I cluster statisticamente significativi sono raggruppatiin strutture ad “Y”o circolarie rappresentano i tessuti potenzialmente patologici. MATERIALI E METODI.Lo spazio metrico adottato è quello di Hamming, esso è in grado di individuare cluster statisticamente significativi sia in pazienti con gliomi omogenei, sia in pazienti con strutture tumorali disomogenee. Per ogni paziente sono state estratte 4 slice dal volume patologico: 2 corrispondenti alle facce superiori inferiori e 2 centrali scelte casualmente. RISULTATI. Per ogni slice analizzata sono stati selezionati i pixel che costituivano i cluster individuati da Iris. I sottinsiemi dei pixel sono stati proiettati sulle immagini originali tracciando delle ROI.Queste maschere sono state confrontate con quelle prodotte dal medico, la qualitá del sistema di segmentazione è stata valutata con il coefficiente di Jaccard, i risultati sono sicuramente incoraggianti (valore medio di Jaccard pari 0.78) che potrebbe essere migliorato eliminando i pochi falsi positivi presenti con tecniche di morfologia matematica o con metodiche a soglia. CONCLUSIONI. Questo approccio basato esclusicavamente su toni di grigio si propone come nuovo strumento per la segmentazioni di immagini Flair. È in atto una sua validazione estensiva su 20 pazienti e sará testato su altre tipologie di immagini diagnostiche e/o su un set piu’ampio di feature. [1] Singh, Gurjeet, Facundo Mémoli, and Gunnar E. Carlsson. "Topological Methods for the Analysis of High Dimensional Data Sets and 3D Object Recognition." SPBG. 2007.

COMPUTER ASSISTED DETECTION IN NEUROIMMAGINI FLAIR E DTI: INDIVIDUAZIONE, SEGMENTAZIONE AUTOMATICA E VOLUMETRIA DEI GLIOMI CEREBRALI Marina Donativi (1,6), Antonella Castellano (2,6), Giorgio De Nunzio (1,6), Gabriella Pastore (3,6), Matteo Rucco (4,6), Antonella Iadanza (2), Marco Riva (5), Lorenzo Bello (5), Andrea Falini (2) 1. Dip. di Matematica e Fisica “Ennio De Giorgi”, Univ. del Salento, Lecce 2. U.O. Neuroradiologia, Ospedale San Raffaele e Univ. Vita-Salute, Milano 3. Istituto di Ricerche Cliniche Ecomedica, Centro di Radioterapia e IGRT, Empoli 4. Univ. di Camerino, School of Science and Technology, Computer Science Division, Camerino 5. U.O. Neurochirurgia, Ist. Clinico Humanitas, Univ. di Milano, Milano 6. Advanced Data Analysis in Medicine, http://adamgroup.it FINALITA'. I sistemi automatici (CAD, Computer Assisted Detection) per la segmentazione e la volumetria dei gliomi cerebrali sono di grande interesse per la valutazione dell’estensione tumorale nella diagnosi, la pianificazione terapeutica e il follow-up: l’uso di un CAD può ridurre la soggettività della diagnosi, aumentandone l’accuratezza. In questo studio le immagini FLAIR e DTI di pazienti con glioma cerebrale sono state elaborate tramite analisi statistica tessiturale 3D per caratterizzare la struttura del tessuto tumorale e sviluppare un tool di segmentazione supervisionata automatica (“GlioCAD”), che fornisca la misura volumetrica delle lesioni e una valutazione degli istogrammi dei livelli di grigio di regioni di interesse [1]. MATERIALI E METODI. 34 pazienti con gliomi di basso e alto grado sono stati sottoposti a RM a 3T con sequenze 3D-FLAIR assiali e DTI (single-shot EPI, b=1000 s/mm2, 32 direzioni). Dalle immagini DTI sono state ottenute le mappe FA, MD, p e q. In ogni mappa sono state segmentate manualmente le regioni patologiche e, a queste ed al tessuto sano controlaterale, è stata applicata l’analisi tessiturale per identificare le caratteristiche (feature) discriminanti. La dimensionalità dello spazio delle feature è stata ridotta tramite Linear Discriminant Analysis (LDA), permettendo la classificazione e la segmentazione automatica del tumore. RISULTATI. Per ogni mappa sono state calcolate sensibilità, specificità e curve ROC, con ottimi risultati (0.90≤AUC≤0.97). Per la fruizione in remoto di GlioCAD è stato realizzato un plugin per OsiriX, che permette caricamento e visualizzazione delle immagini da segmentare, e l’invio al server sul quale risiede GlioCAD. Dopo un tempo di calcolo compatibile con la pratica clinica, le segmentazioni tornano al client e sono visualizzate insieme alle misure volumetriche e agli istogrammi. CONCLUSIONI. GlioCAD si propone come nuovo strumento, basato sull’analisi statistica tessiturale, per la segmentazione automatica dei gliomi cerebrali, la volumetria, e l’analisi quantitativa degli istogrammi nelle regioni di interesse. [1] G. De Nunzio, G. Pastore, M. Donativi, A. Castellano, A. Falini, A CAD system for cerebral gliomas based on texture features in DT-MR images, NIM A, doi:10.1016/j.nima.2010.12.086 (2011)

Computer-Assisted Detection in FLAIR and DT neuroimages: automatic segmentation and volume assessment of cerebral gliomas. Giorgio De Nunzio1,6, Marina Donativi1,6, Antonella Castellano2,6, Gabriella Pastore3,6, Matteo Rucco4,6, Antonella Iadanza2, Marco Riva5, Lorenzo Bello5, Andrea Falini2 (1) Univ. of Salento, Dept. of Mathematics and Physics, and INFN, Lecce (2) U.O. Neuroradiologia, Ospedale San Raffaele e Univ. Vita-Salute, Milano (3) Istituto di Ricerche Cliniche Ecomedica, Centro di Radioterapia e IGRT, Empoli (4) Univ. of Camerino, School of Science and Technology, Computer Science Division, Camerino (5) U.O. Neurochirurgia, Ist. Clinico Humanitas, Univ. di Milano, Milano (6) ADAM srl, Advanced Data Analysis in Medicine, http://adamgroup.it Purpose: tumor cells in cerebral gliomas invade surrounding tissues preferentially along WM tracts, spreading beyond the abnormal area depicted on conventional MR images. Diffusion Tensor Imaging can reveal larger peritumoral abnormalities in gliomas that are not apparent on conventional MRI. We aimed at characterizing pathological vs healthy tissue in FLAIR and DTI datasets by 3D statistical Texture Analysis, developing an automatic segmentation technique for cerebral glioma, hereafter called GlioCAD, especially useful in patient follow-up during chemotherapy, and for preoperative assessment of tumor extension. Methods and materials: thirty-four patients with gliomas were selected. 3T axial 3D-FLAIR, axial 3D-T1w, and DTI (single-shot EPI sequence, b=1000 s/mm2, 32 gradient directions) were acquired. Isotropic and anisotropic maps (FA, MD, p and q) were calculated, and pathological ROIs were manually drawn. 3D texture features were calculated with a sliding window approach in the segmented ROIs and in the contralateral healthy tissue, for CAD-system training. The feature-space dimensionality was reduced by Linear Discriminant Analysis, which allowed tissue classification by simple thresholding. Results: For each map, tumor-classification sensitivity, specificity and ROC curves (0.90≤AUC≤0.97) were calculated, and manual and automatic segmentations were compared by the Jaccard Coefficient, showing good concordance. The CAD system automatically calculated lesion volumes and histograms. With the purpose of allowing remote fruition of GlioCAD, a Graphical User Interface was designed as a plugin for OsiriX, a well-known radiological viewer. Conclusion: GlioCAD is proposed as a new tool, based on statistical textural analysis, for the automatic segmentation and volume assessment of brain gliomas, and for the quantitative analysis of the histograms in the regions of interest.

Objective: To evaluate the serum levels of tumour-associated antigens (TAAs) in patients with SSc and interstitial lung disease (ILD) and to define whether their levels mirror the severity and the progression of lung damage. Methods: Data from 80 SSc patients with ILD were collected at baseline and after 2 years as well as from 40 SSc controls without ILD. The occurrence of any malignancy was recorded. Results: At baseline, an increase of at least one TAA was present in 35 SSc patients with ILD compared with 6 SSc patients without ILD (P < 0.0001); this was associated with lower forced vital capacity (FVC) and higher interstitial and alveolar scores. Levels of carbohydrate antigen 15-3 and carcinoembryonic antigen inversely correlated with FVC and directly correlated with alveolar and interstitial scores and their levels were higher in patients who presented a progression of lung damage after 2 years. During 4 years of follow-up, a malignancy was detected in seven patients who already had an increase of at least one TAA. Values of TAAs increased over time in patients who developed cancer, while their trend remained stable in the others. At multivariate analysis, to have three or more TAAs emerged as a strong independent predictor of the development of malignancies [relative risk 24.1 (95% CI 1.8, 315.0), P = 0.02]. Conclusion: TAAs can be elevated in the sera of SSc patients and correlate with the degree of lung damage, suggesting a role as severity biomarkers. Close follow-up is necessary in SSc patients because of the increased cancer risk overall in patients with increased TAAs.

Background: Arrhythmias are frequent in Systemic Sclerosis (SSc) and portend a bad prognosis, accounting alone for 6% of total deaths. Many of these patients die suddenly, thus prevention and intensified risk-stratification represent unmet medical needs. The major goal of this study was the definition of ECG indexes of poor prognosis. Methods: We performed a prospective cohort study to define the role of 24h-ECG-Holter as an additional risk-stratification technique in the identification of SSc-patients at high risk of life-threatening arrhythmias and sudden cardiac death (SCD). One-hundred SSc-patients with symptoms and/or signs suggestive of cardiac involvement underwent 24h-ECG-Holter. The primary end-point was a composite of SCD or need for implantable cardioverter defibrillator (ICD). Results: Fifty-six patients (56%) had 24h-ECG-Holter abnormalities and 24(24%) presented frequent ventricular ectopic beats (VEBs). The number of VEBs correlated with high-sensitive cardiac troponin T (hs-cTnT) levels and inversely correlated with left-ventricular ejection fraction (LV-EF) on echocardiography. During a mean follow-up of 23.1±16.0 months, 5 patients died suddenly and two required ICD-implantation. The 7 patients who met the composite end-point had a higher number of VEBs, higher levels of hs-cTnT and NT-proBNP and lower LV-EF (p = 0.001 for all correlations). All these 7 patients had frequent VEBs, while LV-EF was not reduced in all and its range was wide. At ROC curve, VEBs>1190/24h showed 100% of sensitivity and 83% of specificity to predict the primary end-point (AUROC = 0.92,p1190/24h had lower LV-EF and higher hs-cTnT levels and, at multivariate analysis, the presence of increased hs-cTnT and of right bundle branch block on ECG emerged as independent predictors of VEBs>1190/24h. None of demographic or disease-related characteristics emerged as predictors of poor outcome. Conclusions: VEBS>1190/24h identify patients at high risk of life-threatening arrhythmic complications. Thus, 24h-ECG-Holter should be considered a useful additional risk-stratification test to select SSc-patients at high-risk of SCD, in whom an ICD-implantation could represent a potential life-saving intervention.

Purpose: Pulmonary embolism (PE), a life-threatening emergency is underdiagnosed because of a non-specific presentation.First-level exams (clinical exhamination, electrocardiography, blood gas analysis and laboratory tests)have low sensitivity and specificity.Clinical prediction rulers (CPRs)such as Wells and Geneva Revised, have been derived from different combinations of these exams. Our aim was to perform a comparison between the two score systems in our population and to derive a new CPR using an Artificial Neural Network (ANN). Methods: We enrolled 755 consecutive outpatients with suspect of PE (351 males, mean age 71±14years) and analyzed 24 clinical,instrumental and laboratoristic variables and Wells and Geneva scores.Logistic regression with ROC curves was used to evaluate the the diagnostic reliability of the scores. To derive the new classifier,the dataset was first split (in supervised classification step)into a train and a test subset containing 2/3 ad 1/3 of the patients' dataset,respectively.To find the optimal configuration of the new classifier we tested two different ANNs:a non-linear feed-forward ANN with back-propagation and a Levenberg-Marquardt network.For both we fixed the topological configurations of the network (one hidden layer,one output neuron)and stressed the system to find the optimal number of neurons in the hidden layer for the best configuration among highest AUC with the highest number of hit in the validation process and the minimum epochs.We repeated this study changing the dimension of the input dataset in two ways:excluding interactively some features or performing the reduction of the dimensionality of the feature space with principal component analysis. The application of the trained ANN to a "map set" gave,for each patient,the probability of belonging to the "pathological" or "healthy" class, obtaining the new CPR.Automatic classifications were compared with the manual ones, calculating the Jaccard coefficient, giving a measure of the quality. The system was implemented in Matlab using Neural Network toolboxes and PRTools. Results: In our population,Wells performed better than Revised Geneva (AUC 0.75%,0.63%,respectively),while our CPR (feed-forward ANN with back-propagation) obtained an average AUC of 0.86% from the train set and Jaccard coefficient 0.86 from the map set.The optimal ANN configuration was with 3 neurons in the hidden layer.The difference among the three ROC curves resulted statistically significant. Conclusions: An ANN-based CPR performs better in the clinical prediction of PE than classical rulers without increasing the number of items required for the analysis.