Advanced Techniques for Face Recognition under Challenging Environments

Automatically recognizing faces captured under uncontrolled environments has always been a challenging topic in the past decades. In this work, we investigate cohort score normalization that has been widely used in biometric verification as means to improve the robustness of face recognition under challenging environments. In particular, we introduce cohort score normalization into undersampled face recognition problem. Further, we develop an effective cohort normalization method specifically for the unconstrained face pair matching problem. Extensive experiments conducted on several well known face databases demonstrate the effectiveness of cohort normalization on these challenging scenarios. In addition, to give a proper understanding of cohort behavior, we study the impact of the number and quality of cohort samples on the normalization performance. The experimental results show that bigger cohort set size gives more stable and often better results to a point before the performance saturates. And cohort samples with different quality indeed produce different cohort normalization performance. Recognizing faces gone after alterations is another challenging problem for current face recognition algorithms. Face image alterations can be roughly classified into two categories: unintentional (e.g., geometrics transformations introduced by the acquisition devide) and intentional alterations (e.g., plastic surgery). We study the impact of these alterations on face recognition accuracy. Our results show that state-of-the-art algorithms are able to overcome limited digital alterations but are sensitive to more relevant modifications. Further, we develop two useful descriptors for detecting those alterations which can significantly affect the recognition performance. In the end, we propose to use the Structural Similarity (SSIM) quality map to detect and model variations due to plastic surgeries. Extensive experiments conducted on a plastic surgery face database demonstrate the potential of SSIM map for matching face images after surgeries.

Development and validation of an echocardiography score for diagnosis of cardiac masses

Background Echocardiography is the cornerstone in the evaluation of cardiac masses and provides accurate characterization. Despite, its accuracy in diagnosis of cardiac masses (CM) remains challenging and, up to date, no validated diagnostic algorithm is validated. Purpose The aim of our study was to evaluate the diagnostic accuracy of echocardiography, to identify the echocardiographic predictors of malignancy and to develop and then validate a multiparametric echocardiographic score that could be used to estimate the likelihood of the histological nature of a CM. Materials and methods The final sample consisted of 273 consecutive patients who had a 2D-echocardiographic evaluation and a histologic diagnosis. Logistic regression was performed to evaluate the ability of echocardiographic findings to discriminate benign versus malignant masses, then a scoring system was developed and validated in a separate test cohort. Results Of the 322 patients initially included in the Bologna Cardiac Masses Registry, 13 with a poor acoustic window, 27 with no histological examination patients and 9 extra-cardiac masses were excluded. In the remaining 273 patients, classical 2-D echocardiogram identified 249 masses with a diagnostic accuracy of 88%. A weighted score [Diagnostic Echocardiographic Mass (DEM) Score] ranging from 0 to 9 was obtained from 6 variables: infiltration, polylobate mass, moderate-severe pericardial effusion. The AUC for the score was 0.965 (95% CI [0.938-0.993]). In a logistic regression analysis using the DEM score as a predictor, the likelihood of malignant CM increased more than 4 times for a 1-unit increase in the score (OR=4.468; 95% CI 2.733-7.304). A score < 3 denoted a high probability of a benign diagnosis, and a score ≥ 5 points corresponded to a higher risk of malignancy. Conclusion 2D-Echocardiography provides a high diagnostic accuracy in identifying cardiac masses and our multiparametric echocardiographic score could be useful to predict the histological nature of cardiac masses.

Valutazione della concordanza e della capacità predittiva degli score prognostici per la definizione di rapida progressione della malattia renale conica nei pazienti con malattia policistica epato renale. Analisi di una coorte multicentrica.

Introduzione: La malattia policistica autosomica dominante (ADPKD) è una causa comune di malattia renale terminale (ESKD). È caratterizzata dallo sviluppo di cisti renali bilaterali che aumentano progressivamente di volume. Il Tolvaptan viene prescritto in base a 3 criteri: volume renale totale (HtTKV) e Mayo Clinic Imaging Class (MCIC), tasso di declino dell'eGFR e al Predicting Renal Outcome in Polycystic Kidney Disease (PROPKD), che combina variabili cliniche e genetiche. In questa coorte multicentrica retrospettiva, l'obiettivo era di valutare e migliorare la concordanza di sensibilità e specificità predittive di MCIC e PROPKD. Metodi: I dati di pazienti adulti affetti da ADPKD sono stati ottenuti da 2 centri di Bologna (B) e Dublino (D). Abbiamo definito RP un calo dell'eGFR ≥3 mL/min/1,73m2/anno su 4 anni (Clinical Score), o classi MCIC 1C-D-E, o punteggio PROPKD da 7 a 9. Per i parametri clinici sono state utilizzate statistiche descrittive. La concordanza tra i punteggi è stata valutata tramite la statistica Kappa. Nelle varianti missenso di PKD1, il punteggio REVEL è stato trattato come una variabile continua; (>0,65 patogeno'). Risultati: Abbiamo valutato 201 pazienti con ADPKD. Il Propkd e il MCIC erano rispettivamente: 90% specifico e 31,3% sensibile; 89,6% sensibile e 28,6% specifico per identificare il calo dell'eGFR. Kappa di Cohen era di 0,025. Il 47,9% (n=143) è risultato concorde. Il punteggio Revel applicato alle mutazioni PKD1NT identifica da 15 a 19 pazienti che potrebbero avere una RP. L'analisi multivariata mostra dati statisticamente significativi per HB (p:0,016), eventi urologici (p: 0,005) e MCIC (p: 0,074). Conclusioni: La concordanza tra i punteggi risulta bassa. Il PROPKD è più selettivo rispetto al Mayo. Tuttavia, il PROPKD permette di identificare alcune RP escluse dall'uso del solo MCIC. L'uso combinato dei punteggi può aumentare la capacità di identificare le RP. Il punteggio REVEL potrebbe migliorare questa concordanza

Medial Artery Calcification (MAC) e Small Artery Disease (SAD) nei pazienti con ischemia critica degli arti inferiori: definizione dei fattori predisponenti, svilup-po di uno score prognostico, definizione istopatologica e valutazione dei risultati dell'arterializzazione delle vene dell'arto inferiore come tecnica di rivascolarizzazione

Questo studio si concentra sull'ischemia critica cronica dell'arto inferiore (CLTI), una patologia globale con gravi complicanze e impatto sociale elevato. Recentemente, la "Medial Artery Calcification" (MAC) è emersa come fattore prognostico significativo nei pazienti con CLTI e malattia grave dei vasi del piede, ma le informazioni sono principalmente retrospettive. Questa tesi esplora la relazione tra MAC e CLTI in tre sezioni. Nella sezione clinica, 248 pazienti sono stati divisi in gruppi MAC per valutare l'impatto prospettico sulla guarigione e sul salvataggio dell'arto. Nella sezione isto-patologica, campioni arteriosi di 26 pazienti sottoposti ad amputazione maggiore sono stati analizzati per comprendere la relazione tra MAC, aterosclerosi e occlusione vascolare. Nella sezione di arterializzazione, 16 pazienti sottoposti all'arterializzazione delle vene del piede (AVP) sono stati esaminati per valutare i risultati clinici prospettici. I risultati della sezione clinica indicano che la presenza di MAC severa è associata a risultati clinici peggiori nei pazienti affetti da CLTI. L'analisi isto-patologica mostra una prevalenza elevata di MAC rispetto all'aterosclerosi, con una associazione importante tra MAC e iperplasia intimale. L'AVP presenta risultati promettenti nei pazienti affetti da CLTI. In conclusione, la MAC influisce sui risultati clinici della CLTI, e l'AVP potrebbe essere una strategia efficace di trattamento.

Developing a radiomic based machine learning model to identify patients with resected non-small-cell lung cancer at high risk of relapse

Background There is a wide variation of recurrence risk of Non-small-cell lung cancer (NSCLC) within the same Tumor Node Metastasis (TNM) stage, suggesting that other parameters are involved in determining this probability. Radiomics allows extraction of quantitative information from images that can be used for clinical purposes. The primary objective of this study is to develop a radiomic prognostic model that predicts a 3 year disease free-survival (DFS) of resected Early Stage (ES) NSCLC patients. Material and Methods 56 pre-surgery non contrast Computed Tomography (CT) scans were retrieved from the PACS of our institution and anonymized. Then they were automatically segmented with an open access deep learning pipeline and reviewed by an experienced radiologist to obtain 3D masks of the NSCLC. Images and masks underwent to resampling normalization and discretization. From the masks hundreds Radiomic Features (RF) were extracted using Py-Radiomics. Hence, RF were reduced to select the most representative features. The remaining RF were used in combination with Clinical parameters to build a DFS prediction model using Leave-one-out cross-validation (LOOCV) with Random Forest. Results and Conclusion A poor agreement between the radiologist and the automatic segmentation algorithm (DICE score of 0.37) was found. Therefore, another experienced radiologist manually segmented the lesions and only stable and reproducible RF were kept. 50 RF demonstrated a high correlation with the DFS but only one was confirmed when clinicopathological covariates were added: Busyness a Neighbouring Gray Tone Difference Matrix (HR 9.610). 16 clinical variables (which comprised TNM) were used to build the LOOCV model demonstrating a higher Area Under the Curve (AUC) when RF were included in the analysis (0.67 vs 0.60) but the difference was not statistically significant (p=0,5147).