Σκοπός της διδακτορικής διατριβής ήταν η αξιολόγηση ενός αυτόματου συστήματος ποσοτικοποίησης της έκτασης της διάμεσης νόσου με απεικονιστικά κριτήρια, όπως την σύγκριση με την ημιποσοτική αξιολόγηση των ακτινολόγων και με την συνεχή αξιολόγηση από ακτινολόγους βασιζόμενη στον χειρωνακτικό σχεδιασμό των παθολογικών περιοχών. Τα αποτελέσματα του αυτοματοποιημένου συστήματος συγκρίθηκαν και με την υποβοηθηση από το σύστημα. Για την οπτικοποίηση της διασύγκρισης των συστημάτων μεταξύ τους χρησιμοποιήθηκαν οι καμπύλες συμφωνίας. Τέλος, μελετήθηκε η συσχέτιση του συστήματος και των ημιποσοτικών εκτιμήσεων των ακτινολόγων με εργαστηριακούς απεικονιστικούς δείκτες, τις πνευμονικές δοκιμασίες. Υλικά και Μέθοδοι Για την εκπόνηση της διατριβής χρησιμοποιήθηκε μια δεξαμενή 47 περιστατικών με νόσο του κολλαγόνου και 4 φυσιολογικών ασθενών. Από αυτήν 14 περιστατικά χρησιμοποιήθηκαν για την ανάπτυξη και εκπαίδευση του συστήματος και 37 περιστατικά επελέγησαν για την αξιολόγηση του συστήματος, με απεικονιστικά πρότυπα θαμβής υάλου (ggo) και δικτυωτού προτύπου (reticular). Το πρωτόκολλο σάρωσης ήταν χαμηλής δόσης απεικόνιση με MDCT 16 ανιχνευτών, που επέτρεπε την λήψη τρισδιάστατων δεδομένων. Αρχικά αναπτύχθηκε το αυτόματο σύστημα σε γλώσσα προγραμματισμού MATLAB (ΜΑΤLAB 7.5 R 2007b) και εκπαιδεύτηκε από τους ακτινολόγους ως προς τα διάφορα στάδια ολοκλήρωσης: (1) Προεπεξεργασία (preprocessing): Τμηματοποίηση πνευμονικών πεδίων και αφαίρεση αγγειακού δέντρου, (2) Ταξινόμηση-Αναγνώριση και χαρακτηρισμός προτύπων διάμεσου ιστού. Το σύστημα αξιολογήθηκε σε δείγμα διαφορετικό από της εκπαίδευσης από την πρώτη δεξαμενή περιστατικών με δείκτες απόδοσης Volume Overlap, ΤPF, FPF. Κατόπιν, το τρισδιάστατο (3D) σύστημα αξιολογήθηκε στο κλινικό δείγμα 37 ασθενών με τις 3D ημιποσοτικές αξιολογήσεις δύο ακτινολόγων (Α΄ Κύκλος Πειραματικού σχεδιασμού). Στην αξιολόγηση αυτή συσχετίστηκαν τα τρισδιάστατα αποτελέσματα ποσοτικοποίησης του CAD ως προς την συνολική έκταση νόσου, την έκταση του ground glass και του reticular προτύπου με τις εκτιμήσεις από δύο ξεχωριστούς έμπειρους ακτινολόγους και του μέσου όρου αυτών ημιποσοτικά στους όγκους των σαρώσεων (περίπου 250 τομές ανά σάρωση ασθενούς) με στατιστική μέθοδο Spearman rank order. Επίσης μελετήθηκαν οι διαφορές μεταξύ των εκτιμήσεων, αναλύθηκαν με Blant –Altman διαγράμματα και boxplots (R1vs. CAD, R2vs.CAD, Rmvs.CAD, R1 vs.R2, R1 vs.R1s). Μετρήθηκε επίσης η συμφωνία ενδο κα μεταξύ των ακτινολόγων. Για τον B΄ κύκλο πειραματικού σχεδιασμού αξιολογήθηκε το 2D σύστημα σε 185 εγκάρσιες τομές (37 σαρώσεις ασθενών) για το ποσοστό της συνολικής έκτασης και της έκτασης του ground glass και του reticular προτύπου. Σαν βάση αναφοράς χρησιμοποιήθηκε η consensus αξιολόγηση δύο έμπειρων ακτινολόγων με σχεδιασμό των παθολογικών περιοχών σε μια τράπεζα σχεδιασμού (Wacom Intuos 3 Tokyo, Japan) με τη βοήθεια μιας γραφικής επιφάνειας διεπαφής (GUI). Επίσης αναλύθηκαν οι ίδιες τομές και αξιολογήθηκαν ημιποσοτικά in consensus από τους ακτινολόγους. Τέλος αξιολογήθηκε στο ίδιο δείγμα το υποβοηθούμενο CAD ως προς την βάση αναφοράς. Συσχετίστηκαν οι 2D εκτιμήσεις του συστήματος και των ακτινολόγων in consensus (1) ημιποσοτικά (CADvsSQRcons) και (2) ποσοτικά με σχεδιασμό των παθολογικών περιοχών pixel-wise (CADvsRref). Οι συσχετίσεις πραγματοποιήθηκαν με Intraclass Correlation Coefficient (ICC) και τα αντίστοιχα 95% Confidence Intervals (CI). Μετρήθηκε επίσης η συμφωνία ενδο κα μεταξύ των ακτινολόγων. Οι διαφορές στην έκταση της νόσου για τις διαφορετικές αξιολογήσεις (CAD - Rref, CAD+Rcons-Rref και SQRcons - Rref) απεικονίστηκαν με Bland-Altman ανάλυση για την συνολική έκταση της νόσου και για τα επιμέρους πρότυπα: το ggo και το reticular. Επίσης υπολογίστηκε η μέση διαφορά (mean difference MD) και το 95% των διαφορών (μέση τιμή±1.96SD), που ονομάζεται όρια συμφωνίας (limits of agreement (LoA95). Για να οπτικοποιηθούν τα αποτελέσματα των αξιολογήσεων της έκτασης νόσου και να διευκολυνθεί η διασύγκριση μεταξύ τους , δημιουργήθηκε μια απεικόνιση με καμπύλες συμφωνίας για κάθε σύστημα αξιολόγησης σε συνάρτηση με την βάση αναφοράς Τέλος συσχετίστηκαν οι τρισδιάστατοι όγκοι των 37 ασθενών σε ότι αφορά (1) την συνολική έκταση της νόσου (2) την έκταση του ggo και (3) την έκταση του reticular όπως εκτιμώνται από το 3DCAD και με την ημιποσοτική αξιολόγηση των ακτινολόγων 3DSQRconsRcons με τις φυσιολογικές παραμέτρους σπειρομέτρησης. ατόπιν συσχετίστηκαν τα 2D δεδομένα στις 185 εγκάρσιες τομές (μέσες τιμές των 5 τομών ανά σάρωση από τους 37 ασθενείς ) για το 2D CAD και την αντίστοιχη ημιποσοτική αξιολόγηση σε 2D SQ . Συγκρίθηκαν οι συσχετίσεις 2DCADvs2DCAD και 3DSQvs2DSQ. με δείκτες συσχέτισης Pearson correlation coefficient (R). Αποτελέσματα Η στατιστική ανάλυση της απόδοσης του συστήματος για την τμηματοποίηση των πνευμονικών πεδίων εκτιμάται ως εξής: (1) Volume overlap: 0.954±0.023, (2)d mean: 1.080±0.364 (3) drms: 1.407±0.735 (4) d max : 4.944±3.492. Η στατιστική ανάλυση της απόδοσης του συστήματος για την τμηματοποίηση του αγγειακού δέντρου και στα δύο πνευμονικά πεδία εκτιμάται ως εξής: (1) Volume overlap: 0.931±0.027, (2 ) ΤPF:0.935±0.036 (3) FPF:0.074±0.03. Η απόδοση του συστήματος στην αναγνώριση και χαρακτηρισμό των παθολογικών προτύπων της διάμεσης νόσου στα περιστατικά που χρησιμοποιήθηκαν για την εκπαίδευση είναι: Για το πρότυπο ground glass: (1) Volume overlap : 0.794±0.038, (2) ΤPF:0.812±0.045, (3) FPF: 0.163±0.017. Για το reticular πρότυπο οι δείκτες απόδοσης είναι: 1) Volume overlap: 0.883±0.037, (2) ΤPF:0.972±0.013 (3) FPF :0.0971±0.012. Οι ανάλογοι δείκτες για την συμφωνία των ακτινολόγων ενδο και μεταξύ- έχουν ως εξής: Για την συμφωνία των ακτινολόγων (ενδο-): Για το πρότυπο ground glass: (1) Volume overlap: 0.47±0.20, (2) ΤPF :0.972±0.01, (3) FPF : 0.0971±0.012. Για το reticular πρότυπο οι δείκτες απόδοσης είναι: (1) Volume overlap : 0.53±0.19, (2)ΤPF: 0.715±0.20, (3)FPF: 1.72±1.05. Για την συνολική έκταση της νόσου, οι αντίστοιχες τιμές είναι: (1) Volume overlap: 0.54 ± 0.18, (2) ΤPF : 0.74 ± 0.18,(3) FPF :1.19±1.85. Για την συμφωνία των ακτινολόγων (μεταξύ-): Για το πρότυπο ground glass: (1) Volume overlap: 0.33±0.22, (2) ΤPF: 0.58±0.27, (3)FPF: 2.48±2,74. Για το reticular πρότυπο οι δείκτες απόδοσης είναι: (1)Volume overlap: 0.51±0.11, (2)ΤPF: 0.72±0.19, (3) FPF:1.36±1.33. Για την συνολική έκταση της νόσου, οι αντίστοιχες τιμές είναι: (1)Volume overlap: 0.51±0.21, (2)ΤPF: 0.69±0.22, (3) FPF: 1.21±0.89. Για τον Α΄ κύκλο πειραματικού σχεδιασμού σε 3D, τα αποτελέσματα δείχνουν στατιστικά σημαντική συσχέτιση της ποσοτικοποίησης του CAD και των ημιποσοτικών εκτιμήσεων των ακτινολόγων (ως βάση αναφοράς), σε ότι αφορά στην συνολική έκταση της νόσου και την έκταση του reticular προτύπου (R=0.949, p=<0.0001, R= 0.915, p=<0.0001, αντίστοιχα), ενώ μέτρια ήταν η συσχέτιση για το ggo πρότυπο (0.806, p=0.0009). Οι ανάλογοι δείκτες για την συμφωνία των ακτινολόγων ενδο και μεταξύ- έχουν ως εξής: Για την συμφωνία των ακτινολόγων (ενδο-): Συνολική έκταση της νόσου: R=0.903, p=<0.0001, έκταση του reticular προτύπου: R= 0.966, p=<0.0001 και έκταση του ggo προτύπου: 0.766, p=<0.0001. Για την συμφωνία των ακτινολόγων (μεταξύ-): Συνολική έκταση της νόσου: 0.838, p=0.0018, έκταση του reticular προτύπου: R= 0.895, p=0.0006 και έκταση του ggo προτύπου: R= 0. 655, p=0.0017. Από την μελέτη των διαφορών όπως απεικονίζονται από τις αναλύσεις Blant Altman και boxplots, οι διαφορές στην έκταση της νόσου μεταξύ ακτινολόγων και CAD (R1 vs.CAD, R2vsCAD, RmvsCAD) και ακτινολόγων μεταξύ τους (R1 vs. R2, R1 vs. R1s) δεν ήταν στατιστικά σημαντικές (two-tailed Wilcoxon signed-rank test, p>0.05). Το εύρος της συσχέτισης κυμαίνεται εντός της μεταβλητής ένδο- και μεταξύ- παρατηρητών, καταδεικνύοντας ένα αξιόπιστο αυτόματο σύστημα ποσοτικοποίησης της διάμεσης νόσου με παρόμοια απόδοση με τους ακτινολόγους. Ωστόσο, η συμπεριφορά των διαφορών CAD vs. R1, CAD vs. R2, CAD vs Rm σε σύγκριση τις διαφορές R1 vs. R2 και R1 vs. R1’ μελετώντας τις median and IQR τιμές δεν είναι όμοια, καταδεικνύοντας ένα αυτόματο σύστημα συμπεριφέρεται με διαφορετικό τρόπο συγκριτικά με την ημιποσοτική εκτίμηση. Οι μέγιστες median and IQR τιμές (3.6% και 32.6%, αντιστοίχως), εμφανίζονται μεταξύ CAD and R2, και αποδίδονται πρωτίστως στην διαφορετική εμπειρία των ακτινολόγων. Τέλος, από τα διαγράμματα Blant Altman παρατηρείται ότι σε όλα τα ζεύγη διαφορών CADvs. R1, CADvsR2, CADvsRm, R1vsR2, R1 vsR1’ οι μεγαλύτερες διαφορές προκύπτουν καθώς αυξάνεται η έκταση της νόσου, ιδίως για έκταση νόσου >20%. Από τα δύο πρότυπα, το reticular πρότυπο παρουσιάζει την μεγαλύτερη μεταβλητότητα της μέσης τιμής. Λαμβάνοντας υπόψιν την διαφρορετική συμπεριφορα στην ποσοτικοποίηση της έκτασης της νόσου μεταξύ του συστήματος CAD και της ημιποσοτικής μεθόδου, αναζητήθηκε μια ακριβέστερη βάση αναφοράς , που στηρίχτηκε στον χειρωνακτικό σχεδιασμό των παθολογικών περιοχών οδηγώντας στον Β’ κύκλο του πειραματικού σχεδιασμού. Για τον B΄ κύκλο πειραματικού σχεδιασμού της διατριβής, το σύστημα παρουσιάζει τόσο με τις ημιποσοτικές όσο και με τις ποσοτικές μετρήσεις, με καλύτερη την συσχέτιση με τις ποσοτικές μετρήσεις που χρησιμοποιήθηκε στην φάση αυτή της διατριβής σαν βάση αναφοράς (ICC =0.809 [0.599-0.894] , 0.851 [0.795-0.891]). Για τα επιμέρους πρότυπα η ποσοτικοποίηση του ground glass από το σύστημα συσχετίζεται λιγότερο με τις εκτιμήσεις των ακτινολόγων ( και με τις δύο μεθόδους, ημιποσοτικά και ποσοτικά) από το reticular. Σε κάθε περίπτωση η ποσοτική αξιολόγηση των ακτινολόγων που θεωρητικά αποτελεί καλύτερη βάση αναφοράς συσχετίζεται σταθερά καλύτερα με το CAD για όλα τα πρότυπα. Η συμφωνία μεταξύ ακτινολόγων είναι παρόμοια και για τις δύο μεθόδους (ICC =0.856 [0.811-0.891], 0.856 [0.806-0.893]). Η ίδια τάση παρατηρείται και για τα δυο πρότυπα (reticular και ground glass). Από την μελέτη των διαφορών όπως απεικονίζονται από τις αναλύσεις Blant Altman, το σύστημα παρουσιάζει ένα μικρό βαθμό υπερεκτίμησης της έκτασης νόσου συγκριτικά με την βάση αναφοράς: Συνολική έκταση νόσου (MD: 2.5%; LoA95: -14.3%, 19.3%), έκταση ground glass: (MD: 1.3%; LoA95: -10.9%, 13.5%) έκταση reticular: (MD: 1.2%; LoA95: -12.7%, 15.1%) και σημαντική συμφωνία με την νέα βάση αναφοράς: 86.5%, 75.1% και 81.6% αντίστοιχα. Οι ημιποσοτικές μετρήσεις παρουσιάζει μεγαλύτερη υπερεκτίμηση με τη βάση αναφοράς: Συνολική έκταση νόσου MD: 9.2%; LoA95: -12.8%, 31.2%, έκταση ground glass: MD: 4.6%; LoA95: -9.8%, 19.0% και έκταση reticular: MD: 4.6%; LoA95: -10.0%, 19.3% και σαφώς μικρότερη συμφωνία με την βάση αναφοράς (69.2%, 70.8%, 70.3%, αντίστοιχα). Τέλος η συμφωνία με την βάση αναφοράς βελτιώνεται ελάχιστα με την χρήση υποβοηθούμενου CAD σε σύγκριση με το CAD σύστημα μόνο του, ιδίως για το ground glass πρότυπο (88.1%, 78.4% and 81.6%, αντίστοιχα). Προτείνεται επίσης στην παρούσα διατριβή η χρήση των καμπυλών συμφωνίας, όπου οπτικοποιείται η διασύγκριση μεταξύ τους, αποδεικνύοντας το υποβοηθούμενο CAD ως το βέλτιστο σύστημα, ακολουθούμενο με μικρή διαφορά από το CAD , ενώ η ημιποσοτικές μετρήσεις διαφέρουν σημαντικά από τα άλλα δύο και από την χειρωνακτική βάση αναφοράς. Oι καμπύλες τονίζουν επίσης την διάσταση των δυο διαφορετικών μεθόδων αξιολόγησης από τους ακτινολόγους ημιποσοτικής-χειρωνακτικού σχεδιασμού, επιβεβαιώνοντας τα αποτελέσματα της πρώτης φάσης της διατριβής. Η εφαρμογή διαφορετικών διαστημάτων διαφορών ±5% ή ±25 % σε συνάφεια με το βήμα ±5% της ημιποσοτικής κλίμακας κατά Desai ή 25% της κλίμακας Likert στην εκτίμηση της έκτασης της νόσου, δείχνουν τα εξής: ‘Οσο πιο λεπτομερές(±5%) είναι το βήμα της ποσοτικοποίησης τόσο μεγαλύτερα τα σφάλματα ποσοτικοποίησης και ιδίως με την ημιποσοτική μέθοδο. Αντίθετα, ενώ για αδρό βήμα 25% οι διαφορετικές μέθοδοι ποσοτικοποίησης συμφωνούν μεταξύ τους και οι μεταξύ τους διαφορές δεν είναι αντιληπτές. Τέλος, το 3D σύστημα συσχετίζεται αρνητικά με όλους τους πνευμονικούς δείκτες για την συνολική έκταση νόσου με με καλύτερη την συσχέτιση με τους δείκτες FEV1. και DLCO ( R=-0.545 p=<0.0001, R=-0.567 p=<0.0001, αντίστοιχα). Για το reticular η συσχέτιση ήταν μέτρια για όλους τους δείκτες με καλύτερη την συσχέτιση FEV1, TLC και DLCO (R=-0.602p=<0.0001, R=-0.615 p=<0.0001 αντίστοιχα), ενώ για το ggo το σύστημα παρουσιάζει μη στατιστικά σημαντική συσχέτιση για όλους τους δείκτες. Tο 2D σύστημα παρουσιάζει μέτρια αρνητική συσχέτιση με τις πνευμονικές δοκιμασίες για την συνολική έκταση νόσου και το reticular πρότυπο, με καλύτερη την μέτρια αρνητική συσχέτιση με τον DLCΟ (R=-0.485, p= 0.002, R=-0.601 p=<0.001 αντίστοιχα). Για το ggo πρότυπο, το 2D σύστημα παρουσιάζει επίσης μη στατιστικά σημαντική συσχέτιση για όλους τους δείκτες, όπως και το 3D σύστημα. Οι ημιποσοτικές μετρήσεις των ακτινολόγων για την συνολική έκταση νόσου 3DSQRconsRcons και 2DSQ παρουσιάζουν στατιστικά σημαντική συσχέτιση, ωστόσο ασθενή μόνο με τον δείκτη DLCO (R=-0.382, p= 0.020, R=-0.398,p=<0.02 αντίστοιχα). Για την 3DSQRcons εκτίμηση στο reticular πρότυπο παρατηρείται στατιστικά σημαντική ασθενής συσχέτιση για όλους τους δείκτες εκτός από τον TLC (FVC: R=-0.331, p= 0.045, FEV1: R=-0.393, p=0.016, DLCO : R=-0.392, p= 0.016). Για την 2DSQ εκτίμηση στο reticular πρότυπο παρατηρείται στατιστικά σημαντική ασθενής έως μέτρια συσχέτιση για όλους τους δείκτες (FVC: R=-0.360, p= 0.029, FEV1: R=-0.411, p=0.012, TLC : R=-0.326, p= 0.049, DLCO : R=-0.485, p= 0.002). Για το ground glass η συσχέτιση είναι μη στατιστικά σημαντική για όλους τους πνευμονικούς δείκτες τόσο για την 3DSQRcons όσο και για την 2DSQ εκτίμηση. Συμπεράσματα Αξιολογήθηκε ένα σύστημα ποσοτικοποίησης της διάμεσης νόσου από CT θώρακος, που βασίζεται σε μεθόδους ανάλυσης υφής, το οποίο ποσοτικοποιεί τα βασικά απεικονιστικά πρότυπα της νόσου, το ggo και το reticular πρότυπο. Αρχικά αξιολογήθηκε σε κάθε βήμα από δύο έμπειρους ακτινολόγους με ικανοποιητική απόδοση. Στη συνέχεια, εφαρμόστηκε ο 3D αλγόριθμος ποσοτικοποίησης σε διαφορετικό κλινικό δείγμα 37 ασθενών, προκειμένου να γίνει και κλινική αξιολόγηση της απόδοσης του με βάση ημιποσοτικές κλίμακες ποσοτικοποίησης, που χρησιμοποιούνται στην βιβλιογραφία, με πολύ καλή συσχέτιση. Με αναλυτικότερη μελέτη προέκυψε μια μεγαλύτερη απόκλιση των τιμών CAD και ημιποσοτικών αξιολογήσεων από τις ημιποσοτικές αξιολογήσεις μεταξύ τους. Το γεγονός αυτό οδήγηση στην περαιτέρω αξιολόγηση του CAD συστήματος με μια χειρωνακτική βάση αναφοράς σε 2D επίπεδο και σύγκριση με τις ημιποσοτικές μετρήσεις. Το σύστημα παρουσίαζε σημαντική συσχέτιση με την βάση αναφοράς και με τις ημιποσοτικές μετρήσεις. Ωστόσο επιβεβαιώθηκε ότι οι ημιποσοτικές μετρήσεις υπολείπονται συγκριτικά με την χειρωνακτική βάση αναφοράς από τους ακτινολόγους στην αξιολόγηση του CAD, όπως είναι φανερό και από τις καμπύλες διασύγκρισης των συστημάτων. Από τη διατριβή αυτή προκύπτει επίσης ένας αριθμός δευτερογενών συμπερασμάτων: Πρώτον, η ποσοτικοποίηση της διάμεσης νόσου παρουσιάζει αυξημένη μεταβλητότητα για όλα τα συστήματα αξιολόγησης , όσο αυξάνεται η έκταση της νόσου. Δεύτερον, από τα δύο πρότυπα, η ποσοτικοποίηση του ggo προτύπου παρουσίαζε τις μεγαλύτερες δυσκολίες. Τρίτον, η αξιολόγηση με υποβοηθηση CAD βελτιώνει ελάχιστα το αποτέλεσμα του συστήματος, και κυρίως διορθώνει το ggo πρότυπο. Τέταρτον, το σύστημα συσχετίζεται με τις πνευμονικές δοκιμασίες, περισσότερο από τις ημιποσοτικές εκτιμήσεις των ακτινολόγων. Ιδίως το reticular πρότυπο μπορεί να χρησιμοποιηθεί από μόνο του ως δείκτης πνευμονικής δυσλειτουργίας. Συνοψίζοντας, το σύστημα CAD που αναπτύχθηκε είναι ένα αξιόπιστο εργαλείο ποσοτικοποίησης, το οποίο μπορεί να χρησιμοποιηθεί ως βιολογικός δείκτης σταδιοποίησης της διάμεσης νόσου. / The aim of this thesis was the evaluation of an automated system in 3D quantification of interstitial lung disease extent in CT compared to semiquantitative scoring method and with continuous extent assessments by radiologists based on manual segmentations of abnormal areas. The CAD system output was also compared to extent assessments by radiologists assisted by CAD. To facilitate intercomparison among different evaluation assessments curve representations of extent assessments is proposed. Finally the system output semiquantitative assessment by radiologists were correlated with laboratory markers of ILD, meaning pulmonary function tests (PFTs) indexes. Materials and Methods A dataset of 47 patients with interstitial lung disease secondary to collagen vascular disease were recruited for this thesis as well as 4 normal patients. Out of the dataset, 14 cases were used to train the system and 37 cases were selected to evaluate the system. The above cases presented with ground glass and reticular patterns. A low-dose MDCT scanner (16 detectors) protocol was utilized to obtain volumetric system output. Initially, the automated quantification system was developed using MATLAB (ΜΑΤLAB 7.5 R 2007b) and was trained by radiologists in different development stages: (1) Preprocessing: Segmentation of lung fields and vessel tree segmentation, (2) Identification and Classification of interstitial lung disease. The evaluation performance of the system was evaluated in terms of volume overlap, ΤPF, FPF. Following, the system was evaluated in a case sample of 37 patient scans with semiquantitative scoring by two experienced radiologists (A’ cycle of experimental design). In this evaluation, correlation between 3D system output and 3D semiquantitative evaluations of volumetric scans (approximately 200 slices per patient scan) was performed by two expert radiologists and average of the two evaluations, regarding total disease, ground glass and reticular pattern extent, using Spearman rank order. The differences in extent assessment (R1vsCAD, R2vsCAD, RmvsCAD, R1vsR2, R1vsR1s) were also analyzed using Blant–Altman plots and box plots. Statistic evaluation was performed using two-tailed Wilcoxon signed-rank test for paired data. Intra and interobserver agreement was also measured. At the B’ cycle of experimental design the 2D system output was evaluated, out of 185 axial slices (37 patient scans) regarding total disease, ground glass and reticular pattern extent, As reference standard in consensus evaluation by two expert radiologists was used by drawing disease area segments on a drawing tablet (Wacom Intuos 3 Tokyo, Japan), utilizing a home-developed graphical-user interface (GUI). The same slices were also assessed semiquantitatively by radiologists in consensus. Finally, the same slices were assessed with assisted CAD and compared to reference standard. The 2D CAD output was correlated with semiquantitative evaluation by radiologists (CADvsSQRcons) and quantitatively by drawing disease area segments pixel-wise (CADvsRref). Correlations were performed with Intraclass Correlation Coefficient (ICC) and 95% Confidence Intervals (CI). Intra and interobserver agreement was also measured. Differences in extent assessment for different evaluations (CAD- Rref, CAD+Rcons -Rref and SQRcons -Rref) were plotted using Bland-Altman analysis were assessed for total disease, ground glass and reticular pattern extent. The mean difference (MD) and 95% of the differences (mean±1.96 standard deviation), called limits of agreement (LoA95), were calculated. A curve representation is proposed to visualize agreement and facilitate agreement inter-comparison for various evaluations, as a function of reference. Finally, volumetric system output of 37 patient scans were correlated regarding total disease, ground glass and reticular pattern extent (3DCAD) and volumetric semiquantitative evaluation by radiologists (3DSQRcons) were correlated with physiologic parameters, pulmonary function tests, in terms of Pearson correlation (R). The same correlations were performed for 185 axial slices (average of 5 slices per patient scan) for 2DCAD and 2DSQ. Comparison of 3DCAD vs2D CAD and 3DSQvs2DSQ was also performed. Results Performance evaluation of segmentation of lung fields was measured: (1) Volume overlap: 0.931±0.027, (2) ΤPF: 0.935±0.036 (3) FPF: 0.074±0.03. Performance evaluation of segmentation of vessel tree was measured: (1) Volume overlap: 0.931 ±0.027, (2) ΤPF: 0.935±0.036 (3)FPF: 0.074±0.03. Performance evaluation in identification and characterization of ILD patterns was measured: For ground glass pattern: (1) Volume overlap: 0.794±0.038, (2) ΤPF: 0.812±0.045 (3) FPF: 0.163±0.017. For reticular patterns: (1) Volume overlap: 0.883 ± 0.037, (2) ΤPF: 0.972±0.013(3) FPF: 0.0971±0.012. Performance evaluation of intraobserver agreement was: For ground glass pattern: (1) Volume overlap : 0.47±0.20, (2) ΤPF: 0.972±0.013 (3) FPF: 0.0971±0.012. For reticular patterns: (1) Volume overlap: 0.53±0.19, (2) ΤPF: 0.715±0.20 (3) FPF: 1.72±1.05. For total disease extent: (1) Volume overlap: 0.54±0.18, (2) ΤPF: 0.74±0.18(3) FPF: 1.19±1.85. Performance evaluation of interobserver agreement was: For ground glass pattern: (1) Volume overlap: 0.33±0.22, (2) ΤPF: 0.58±0.27, (3) FPF: 2.48±2,74. For reticular patterns: (1) Volume overlap: 0.51±0.11, (2) ΤPF: 0.72± .19,(3) FPF: 1.36±1.33. For total disease extent: (1)Volume overlap: 0.51±0.21, (2)ΤPF: 0.69±0.220, (3)FPF: 1.21±0.89. The A’ cycle of experimental design in 3D quantification, results show statistically significant correlation of CAD quantification output and semiquantitative evaluation by radiologists (reference standard), regarding total and reticular disease extent (R=0.949, p=<0.0001, R= 0.915, p=<0.0001, respectively), while moderate correlation regarding ground glass pattern (R=0.806, p=0.0009). Intra and interobserver agreement were as follows: For Intra observer agreement: For total disease extent: R=0.903, p=<0.0001, reticular pattern: R=0.966, p=<0.0001 and ground glass pattern and R : 0.766, p=<0.0001. For Interobserver agreement: For total disease extent: R=: 0.838, p=0.0018, reticular pattern: R= 0.895, p=0.0006 and ground glass pattern: R= 0. 655, p=0.0017. Observed differences in extent assessment as plotted at Blant Altman και boxplot representation, between radiologists and CAD (R1 vs. CAD, R2 vs. CAD, Rm vs. CAD), radiologists themselves (R1 vs. R2, R1 vs. R1s) were not statistically significant (two-tailed Wilcoxon signed-rank test for paired data, p>0.05). Results indicate that the CAD tool analyzed demonstrates similar performance to radiologists semiquantitative assessment, ranging within inter- και intra- observer variation and can be considered as a reliable independent reader of lung abnormalities in ILD. However, differences of extent assessment between radiologists and CAD are (CADvsR1, CADvsR2, CADvsRm ) differ significantly to those obtained between radiologists (R1vsR2 ) and radiologist second opinion (R1vsR1s) in terms of median and IQR values. This implies that the automated system compared to the semiquantitative radiologist assessment, does not behave in an identical manner, as indicated by decreased variability of inter- and intraobserver data. The maximum median and IQR values (3.6% and 32.6%, respectively), were obtained between CAD and R2, probably attributed to experience differences between experts. Finally. Blant Altman plots indicate that in all differences in extent CADvsR1, CADvsR2, CADvsRm, R1vsR2, R1vsR1’ increasing differences occur in increasing extent assessments, especially> 20%. Out of the two patterns, reticular presents greater variability of mean values. Considering the different behavior in extent assessment between CAD system and semiquantitative scoring, we considered a more accurate reference standard based on manual drawing of abnormal areas, leading to the B’ cycle of experimental design. In B’ cycle of experimental design, the system shows significant correlation to semiquantitative and quantitative extent assessments (ICC =0.809 [0.599-0.894], 0.851 [0.795-0.891]). Correlation to quantitative extent assessments was slightly better and was used as a reference standard for this phase of thesis. For the constituent patterns quantification of ground glass by CAD system correlates less with radiologists assessments (both semi quantitatively and quantitatively) than reticular pattern. In every case quantitative evaluation by radiologists correlates better in all patterns with CAD system. Interobserver agreement is similar for both evaluation methods (ICC =0.856 [0.811-0.891], 0.856 [0.806-0.893]). The same trend is observed for both patterns (reticular and ground glass). Analyzing the differences as plotted by Blant Altman, the system depicts a small degree of overestimation of extent assessment as compared to reference standard. Total disease extent: (MD: 2.5%; LoA95: -14.3%, 19.3%), ground glass extent (MD: 1.3%; LoA95: -10.9%, 13.5%), reticular pattern extent (MD: 1.2%; LoA95: -12.7%, 15.1%). The system shows substantial agreement to pixel-wise (reference): 86.5%, 75.1% και 81.6%, for total lung disease, ground glass and reticular pattern, respectively. Semiquantitative scoring demonstrates higher degree of disease extent overestimation to reference: Total disease extent: MD: 9.2%; LoA95: -12.8%, 31.2%, ground glass extent: MD: 4.6%; LoA95: -9.8%, 19.0% and reticular extent: MD: 4.6%; LoA95: -10.0%, 19.3%) and achieves significantly lower agreement to reference (69.2%, 70.8%, 70.3%, respectively). Finally, agreement to reference achieved by assisted CAD, is slightly improved compared to CAD alone, especially for ground glass pattern (88.1%, 78.4% and 81.6%, respectively). The utilization of curve representation of the degree of agreement is proposed in this thesis, in an effort to visualize agreement of extent assessments and to facilitate agreement inter-comparison. These curves indicate assisted CAD as the best system, presenting with the higher agreement to reference, with CAD alone nearby, while semiquantitative scoring differs significantly from the manually drawn reference. CAD system and assisted CAD. These curves depict distinctively the differences in agreement between the two evaluation approaches by radiologists: semiquantitative and pixel-wise, confirming the observations of the first part of the thesis. On these curves difference interval ±5% ή ±25 % in extent assessment can be applied, according to 5% step used in semiquantitative scoring by Desai et al. or 25% in Likert scaling. Τhe more detailed (±5%) the quantification step is the greater the quantification errors are. Οn the contrary for wider step 25% the different the differences among them are not perceivable Finally, comparison of volumetric system output (3D) with PFTs yields strong negative correlations for total disease with all indexes, with better correlation with FEV1 and DLCO (R=-0.545 p=<0.0001, R=-0.567 p=<0.0001, respectively). For reticular pattern, correlation was moderate to all PFT indexes with better correlation with FEV1 and DLCO (R=-0.602, p=<0.0001, R=-0.615 p=<0.0001, respectively). For ground glass pattern there is no statistical correlation to neither of PFT indexes. For 2D system output, negative correlation was shown to PFTs for total disease and reticular pattern. Best moderate negative correlation was shown with DLCΟ (R=-0.485, p= 0.002, R=-0.601 p=<0.001 respectively). For ggo pattern, 2D system output showed no statistically significant correlation to all indexes, as in 3D system output. Semiquantitative scoring 3DSQRconsRcons and 2DSQ showed for total disease statistically significant negative correlations to PFTs, albeit weak, only with index DLCO (R=-0.382, p= 0.020, R=-0.398p=<0.02, respectively).. For 3DSQRcons semiquantitative scoring, reticular pattern showed statistically significant weak negative correlation to PFTs (not included to TLC) (FVC: R=-0.331, p= 0.045, FEV1: R=-0.393, p=0.016, DLCO : R=-0.392, p= 0.016). For 2DSQ semiquantitative scoring, reticular pattern showed statistically significant weak to moderate negative correlation to all PFTs (FVC: R=-0.360, p= 0.029, FEV1: R=-0.411, p=0.012, TLC : R=-0.326, p= 0.049, DLCO: R=-0.485, p= 0.002). For ground glass there was no statistically significant correlation for all indexes regarding 3DSQRcons and 2DSQ valuation. Conclusions In conclusion, an automated quantification system for ILD extent in CT was assessed, based on texture features, quantifying the two major imaging patterns of the disease, ggo and reticular pattern. Initially the system was evaluated by two experiences radiologists with performance within interobserver variability. Accordingly, the 3D system output was evaluated in a different dataset of 37 patients in order to evaluate the results with semiquantitative scoring published in literature. The correlation was very good. Further analysis of differences showed greater deviation of CAD and semiquantitative quantification, compared to semiquantitative quantification by radiologists (interobserver agreement). This remark led to further evaluation of CAD with a reference standard in 2D, based on manual drawings by radiologists and comparison with semiquantitative quantification. The 2D system output showed significant correlation to the reference standard and semiquantitative scoring. However semiquantitative quantification are less accurate in assessment of CAD quantification systems than quantitative quantification by manual drawing by radiologists. This is depicted in curve representation of different scoring systems compared to reference. From this thesis a number of secondary conclusion are drawn: First, in all systems, quantification variability augments with augmentation of extent. Second, between the two patterns analyzed quantification of ground glass is more difficult. Moreover, assisted CAD slightly improves the system output and mostly the ground glass pattern is corrected. Finally, the system correlates better to PFTs than semiquantitative quantification by radiologists. Especially reticular pattern by itself can be used as a reliable marker of pulmonary dysfunction. Overall, the system analysed is a useful disease extent quantification tool, than could act as a biomarker for staging and follow up of interstitial lung disease.

Diacerhein (DAR) is a slow acting symptomatic drug used to treatment of osteoarthritis. It is available in the Brazilian market as capsules and compounded capsules. There are no official methods for DAR analysis in any of the pharmacopeia. In the present work, methods for the quantification and dissolution evaluation of the drug in capsules and compounded capsules were developed and validated. Spectrophotometry and high-performance liquid chromatography (HPLC) were the methods used for drug determination. In the spectrophotometric method the DAR can be quantified at 277 as well as 502 nm, using 0.1N NaOH as diluent. The HPLC analyses were performed on a C18 column, using a mobile phase composed of triethylamine 0.1%:acetonitrile (65:35, v/v; adjusted to 7.5 with orthophosphoric acid) and UV detection at 250 nm. The methods showed good linearity (r>0.99), precision (RSD<2%) and accuracy (>99%) and the results obtained were not statistically different (P=0.05). The optimization of dissolution test conditions for in vitro quality control of DAR in capsules was evaluated. The use of 900 mL of sodium phosphate buffer at 37.0 ± 0.5 ºC, basket as apparatus, at 100 rpm rotate speed and 30 minutes of test provided satisfactory results for tested products. The percent dissolution of DAR in the established condition was more than 80% and the different evaluated products presented good dissolution efficiency (>85%). The spectrophotometric method that was validated to evaluate the dissolution testing showed to be specific, with no interference of the placebo or shell capsules in the quantification of DAR, linear (0.27 6.66 μg mL-1; r>0.99), precise (RSD<5%) and accurate (>97%). The drug showed satisfactory stability in the selected dissolution medium. / A diacereína (DAR) é um fármaco antiartrósico sintomático de ação lenta. No mercado brasileiro encontra-se disponível na forma de cápsulas e cápsulas manipuladas. Não existem monografias para análise de DAR em nenhuma farmacopéia. No presente trabalho, métodos para quantificação e avaliação da dissolução do fármaco em cápsulas foram desenvolvidos e validados. Os métodos usados para a quantificação do fármaco foram a espectrofotometria e a cromatografia líquida de alta eficiência (CLAE). No método espectrofotométrico DAR pode ser quantificada em 277 ou em 502 nm, utilizando NaOH 0,1N como diluente. As análises por CLAE foram realizadas em coluna C18, utilizando fase móvel composta de trietilamina 0,1%: acetonitrila (65:35, v/v; com pH ajustado para 7,5 com ácido ortofosfórico) com detecção no UV em 250 nm. Os métodos mostraram boa linearidade (r>0,99), precisão (DPR<2%) e exatidão (>99%) e os resultados obtidos não apresentaram diferença estatística significante (P=0,05. A otimização das condições para o teste de dissolução in vitro para DAR em cápsulas foi avaliada. As condições que forneceram resultados satisfatórios para os produtos testados foram 900 mL de tampão fosfato de sódio pH 7,0 com 0,75% de lauril sulfato de sódio, a 37,0 ± 0,5 ºC, aparato cesta, 100 rpm e 30 minutos de teste. A liberação de DAR na condição estabelecida foi superior a 80% e os diferentes produtos analisados demonstraram boa eficiência de dissolução (>85%). O método espectrofotométrico que foi validado para avaliar a dissolução mostrou ser específico, não havendo interferência de excipientes ou do invólucro das cápsulas na quantificação de DAR, linear (0,27 6,66 μg/mL; r> 0,99), preciso (DPR < 5%) e exato (>97%). O fármaco mostrou estabilidade satisfatória no meio de dissolução selecionado.

Made available in DSpace on 2015-05-14T12:59:24Z (GMT). No. of bitstreams: 1 parte1.pdf: 2506854 bytes, checksum: 7b83f48e46031c35dc85c825fdbe1a10 (MD5) Previous issue date: 2010-08-30 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / Resveratrol (cis and trans) and quercetin are important components in ensuring the quality of wine because of the diversity of biological activities that they have. Studies quantifying these constituents in Brazilian wines are rarely reported in the literature, especially in wines from the Vale Submédio do São Francisco. The objectives of this work were to develop and validate a quantification method for cis-resveratrol, trans-resveratrol, total resveratrol and quercetin by HPLC-UV in red wines, to evaluate and correlate the levels of total phenolics by Folin-Ciocalteu method with the anti-radical activity determined usingm the DPPH radical method and to determine the potential for classification of wines from their chromatographic profiles. In the study 45 samples of red wines were analysed. From these, 30 samples were from Brazilian wines:15 of the Vale Submédio do São Francisco, 11 of Serra Gaucha and 4 of Campanha Gaucha. There were 15 samples of imported wines: 5 from Argentina, 5 from Chile and 5 from French. The method developed for quantification of resveratrol (cis and trans) and quercetin was validated andused in the quantification. The results showed that Brazilian wines have about nine times more cis-resveratrol than trans-resveratrol and this relationship is even more pronounced in the Vale do São Francisco, with about 17 times more cis-resveratrol than trans-resveratrol in red wines from this region. The French wines presented quercetin in an amount above the rest of the wine samples. It was also observed that the wines from the Vale do São Francisco had higher levels of total resveratrol, significantly higher than imported wines, with p=0.0381. The total phenolics content was markedly higher in the wines from the Vale do São Francisco in relation to wines from the South, with p=0.001. The study of anti-radical activity against the DPPH radical showed that all wines are active, with EC50<130 μg/mL. In the study of the chromatographic profiles of the wines four distinct groupings among the wines were found, with 87% of the samples from the Vale do São Francisco, 80% of Chileans wines, 60% of the Serra Gaucha wines and 55% of French wines. Taken toghether the results show that Brazilian wines presented high level of phenolics, especially the wines from Vale do Submédio do São Francisco. This is translated into high levels of bioactive substances such as resveratrol (cis and trans) and quercetin, as well the high antioxidant activity. The high intensity of solar light in Brazil, mainly in the Vale do São Francisco, may have been the main factor contributing to the high content of these constituents, thus contributing to the high degree of grouping shown by the wines of this region using multivariate analysis. / O resveratrol (cis e trans) e quercetina são componentes de grande importância na garantia da qualidade do vinho, devido à diversidade de atividades biológicas que eles possuem. Estudos de quantificação destes constituintes em vinhos brasileiros são pouco relatados na literatura, especialmente nos vinhos do Vale Submédio do São Francisco. Os objetivos deste trabalho foram desenvolver e validar um método de quantificação do cis-resveratrol, trans-resveratrol, resveratrol total e quercetina por CLAE-UV/Vis em vinhos tintos, avaliar e correlacionar os teores de fenólicos totais pelo método de Folin-Ciocalteu com a atividade anti-radicalar, frente ao radical DPPH e determinar o potencial de classificação dos vinhos a partir dos perfis cromatográficos. Para execução do trabalho foram analisadas 45 amostras de vinhos tintos, sendo 30 brasileiros das regiões do Vale Submédio do São Francisco (15), Serra Gaúcha (11) e Campanha Gaúcha (4), e 15 importados: argentinos (5), chilenos (5) e franceses (5). O método desenvolvido para quantificação de resveratrol isômeros (cis e trans) e quercetina apresentou dados e características suficientemente bons para ser validado e utilizado no processo de quantificação. Os resultados obtidos demonstraram que os vinhos brasileiros possuem cerca de 9 vezes mais cis-resveratrol que trans-resveratrol, esta relação é ainda mais pronunciada na região do Vale Submédio do São Francisco, sendo cerca de 17 vezes mais cis do que trans. Os vinhos franceses apresentaram quercetina em quantidade superior ao restante dos conjuntos amostrais de vinhos. Foi observado também que os vinhos do Vale Submédio do São Francisco possuem teores de resveratrol total, significativamente superiores aos vinhos importados, com p=0,0381. O teor de fenólicos totais foi marcadamente superior nos vinhos do Vale Submédio do São Francisco em relação aos vinhos da região Sul, com p=0,001. O estudo de atividade anti-radicalar frente ao radical DPPH demonstrou que todos os vinhos são ativos, com CE50 <130 μg/mL. No estudo do perfil cromatográfico dos vinhos foram determinados quatro agrupamentos distintos, sendo 87 % entre os vinhos do Vale Submédio do São Francisco, 80 % entre os chilenos, 60 % entre os da Serra Gaúcha e 55 % entre os vinhos franceses. Os estudos realizados com os vinhos revelaram resultados que posicionam os vinhos brasileiros, especialmente os vinhos da região do Vale Submédio do São Francisco, em um elevado patamar em termos de qualidade. Isto é corroborado pelos elevados teores de substâncias bioativas tais como o resveratrol (isômeros cis e trans) e quercetina, bem como o elevado teor de fenólicos totais e alta atividade anti-radicalar. A alta intensidade de insolação no Brasil, principalmente na região do Vale Submédio do São Francisco, pode ter sido o principal fator a contribuir para elevação do teor destes constituintes, contribuindo, desta forma, para um elevado grau de agrupamento dos vinhos desta região por análise multivariada.

Until recent times, weather forecasts were deterministic in nature. For example, a forecast might state ``The temperature tomorrow will be $20^\circ$C.'' More recently, however, increasing interest has been paid to the uncertainty associated with such predictions. By quantifying the uncertainty of a forecast, for example with a probability distribution, users can make risk-based decisions. The uncertainty in weather forecasts is typically based upon `ensemble forecasts'. Rather than issuing a single forecast from a numerical weather prediction (NWP) model, ensemble forecasts comprise multiple model runs that differ in either the model physics or initial conditions. Ideally, ensemble forecasts would provide a representative sample of the possible outcomes of the verifying observations. However, due to model biases and inadequate specification of initial conditions, ensemble forecasts are often biased and underdispersed. As a result, estimates of the most likely values of the verifying observations, and the associated forecast uncertainty, are often inaccurate. It is therefore necessary to correct, or post-process ensemble forecasts, using statistical models known as `ensemble post-processing methods'. To this end, this thesis is concerned with the application of statistical methodology in the field of probabilistic weather forecasting, and in particular ensemble post-processing. Using various datasets, we extend existing work and propose the novel use of statistical methodology to tackle several aspects of ensemble post-processing. Our novel contributions to the field are the following. In chapter~3 we present a comparison study for several post-processing methods, with a focus on probabilistic forecasts for extreme events. We find that the benefits of ensemble post-processing are larger for forecasts of extreme events, compared with forecasts of common events. We show that allowing flexible corrections to the biases in ensemble location is important for the forecasting of extreme events. In chapter~4 we tackle the complicated problem of post-processing ensemble forecasts without making distributional assumptions, to produce recalibrated ensemble forecasts without the intermediate step of specifying a probability forecast distribution. We propose a latent variable model, and make a novel application of measurement error models. We show in three case studies that our distribution-free method is competitive with a popular alternative that makes distributional assumptions. We suggest that our distribution-free method could serve as a useful baseline on which forecasters should seek to improve. In chapter~5 we address the subject of parameter uncertainty in ensemble post-processing. As in all parametric statistical models, the parameter estimates are subject to uncertainty. We approximate the distribution of model parameters by bootstrap resampling, and demonstrate improvements in forecast skill by incorporating this additional source of uncertainty in to out-of-sample probability forecasts. In chapter~6 we use model diagnostic tools to determine how specific post-processing models may be improved. We subsequently introduce bias correction schemes that move beyond the standard linear schemes employed in the literature and in practice, particularly in the case of correcting ensemble underdispersion. Finally, we illustrate the complicated problem of assessing the skill of ensemble forecasts whose members are dependent, or correlated. We show that dependent ensemble members can result in surprising conclusions when employing standard measures of forecast skill.

abstract: Titanium dioxide (TiO2) nanomaterial use is becoming more prevalent as is the likelihood of human exposure and environmental release. The goal of this thesis is to develop analytical techniques to quantify the level of TiO2 in complex matrices to support environmental, health, and safety research of TiO2 nanomaterials. A pharmacokinetic model showed that the inhalation of TiO2 nanomaterials caused the highest amount to be absorbed and distributed throughout the body. Smaller nanomaterials (< 5nm) accumulated in the kidneys before clearance. Nanoparticles of 25 nm diameter accumulated in the liver and spleen and were cleared from the body slower than smaller nanomaterials. A digestion method using nitric acid, hydrofluoric acid, and hydrogen peroxide was found to digest organic materials and TiO2 with a recovery of >80%. The samples were measured by inductively coupled plasma-mass spectrometry (ICP-MS) and the method detection limit was 600 ng of Ti. An intratracheal instillation study of TiO2 nanomaterials in rats found anatase TiO2 nanoparticles in the caudal lung lobe of rats 1 day post instillation at a concentration of 1.2 ug/mg dry tissue, the highest deposition rate of any TiO2 nanomaterial. For all TiO2 nanomaterial morphologies the concentrations in the caudal lobes were significantly higher than those in the cranial lobes. In a study of TiO2 concentration in food products, white colored foods or foods with a hard outer shell had higher concentrations of TiO2. Hostess Powdered Donettes were found to have the highest Ti mass per serving with 200 mg Ti. As much as 3.8% of the total TiO2 mass was able to pass through a 0.45 um indicating that some of the TiO2 is likely nanosized. In a study of TiO2 concentrations in personal care products and paints, the concentration of TiO2 was as high as 117 ug/mg in Benjamin Moore white paint and 70 ug/mg in a Neutrogena sunscreen. Greater than 6% of Ti in one sunscreen was able to pass through a 0.45 um filter. The nanosized TiO2 in food products and personal care products may release as much as 16 mg of nanosized TiO2 per individual per day to wastewater. / Dissertation/Thesis / M.S. Engineering 2011

Cette thèse porte sur les pratiques et les usages numériques de quantification de soi. Le mouvement Quantified Self est apparu initialement en 2007 dans la Silicon Valley, mais en quelques années ces pratiques ont rapidement évolué pour converger vers les technologies numériques en santé. S’il ressort de la littérature scientifique et académique qu’elles constituent une forme contemporaine de biopouvoir (Lupton, 2016) et qu’elles sont porteuses de nombreux espoirs dans le domaine de la santé, elles ne sont pourtant pas questionnées, ni du point de vue des mutations anthropologiques qu’elles introduisent dans le couplage entre organisme physiologique et données numériques (Simondon, 1958 ; Boullier, 2011 ; Sadin, 2013), ni du point de vue des modèles de conception sous-jacents aux technologies de quantification de soi, essentiellement fondées sur des approches comportementales, privilégiant la persuasion plutôt que la signification. Ce manque de réflexion soulève de nombreuses questions d’ordre éthique quant à la manière de concevoir des dispositifs numériques, en particulier lorsqu’il s’agit de la santé des individus (Lupton, 2013 ; 2016). Dans cette perspective, cette thèse poursuit un double objectif. Le premier est d’apporter un éclairage compréhensif sur les pratiques numériques de quantification de soi. Le second se rapporte à l’instrumentation de ces nouveaux objets technologiques et à leur modélisation en amont de leur conception. Pour ce faire, nous avons choisi le modèle Learning by expanding d’Engeström (1999, 2014) qui permet d’envisager la conception sous l’angle de la médiation. / This thesis concerns the digital practices and uses of self-quantification. The quantified-self movement first appeared in 2007 in Silicon Valley, but in a few years these practices have evolved rapidly to converge on digital health technologies. If it appears from the scientific and academic literature, that they constitute a contemporary form of biopower (Lupton, 2016) and that they carry many hopes in the field of health, they are however not questioned, neither from the perspective of the anthropological changes they introduce in the coupling between body and physiological digital data (Simondon, 1958, Boullier, 2011; Sadin, 2013), nor from the self-quantification technologies models point of view mainly based on behaviorist approaches, favoring persuasion rather than meaning. This lack of thinking raises many ethical questions about how to design a technology, especially when it comes to the health of individuals (Lupton, 2013; 2016).In this perspective, this thesis pursuit a dual purpose. The first is to provide a comprehensive perspective on the digital self-quantification practices. The second objective relates to the instrumentation of these new technological artifacts and to their modeling. To do this, we chose the approach the “Learning by Expanding model” of Engeström (1999, 2014), to model the digital mediations processes involved in the self-quantification practices.

No description available.

Oestrogenic chemicals have become increasingly associated with health effects in wildlife populations and humans. Transgenic animal models have been developed to understand the mechanisms by which these oestrogenic chemicals alter hormonal signalling pathways and how these alterations can lead to chronic health effects. The use of highly informative transgenic animal models will also result in better use and potential reduction of intact animals used in animal testing in line with the principles of the 3Rs. In this thesis work, two novel oestrogen responsive transgenic zebrafish models have been generated to investigate the effects of oestrogenic chemicals, identify their tissue targets and better understand the temporal dynamics of these responses. Both models express the pigment-free ‘Casper’ (a mutant line lacking skin pigment) phenotype, which facilitate identification of responding target tissues in the whole fish in all fish life stages (embryos to adults). The oestrogen response element green fluorescent (ERE-GFP)-Casper model was generated by crossing an established ERE-GFP line with the skin pigment free Casper line. The model generated is highly sensitive to oestrogenic chemicals, detecting responses to environmentally relevant concentrations of EE2, bisphenol A (BPA), genistein and nonylphenol. Use of the ERE-GFP- Casper model shows chemical type and concentration dependence for green fluorescent protein (GFP) induction and both spatial and temporal responses for different environmental oestrogens tested. A semi-automated (ArrayScan) imaging and image analysis system was also developed to quantify whole body fluorescence responses for a range of different oestrogenic chemicals in the new transgenic zebrafish model. The zebrafish model developed provides a sensitive and highly integrative system for identifying oestrogenic chemicals, their target tissues and effect concentrations for exposures in real time and across different life stages. It thus has application for chemical screening to better direct health effects analysis of environmental oestrogens and for investigating the functional roles of oestrogens in vertebrates. The second model generated was an ERE-Kaede-Casper line developed via crossing of the ERE-GFP-Casper line and a UAS-Kaede line and screening subsequent generations for a desired genotype and homozygous expression of the transgenes. Kaede is a photoconvertible fluorescent protein that initially fluoresces green in colour and can be permanently converted to red fluorescence upon short exposure to UV light. The model has a silenced skin pigmentation and high sensitivity to oestrogenic chemicals comparable with the previously developed ERE-GFP-Casper model. Use of this model has identified windows of tissue-specific sensitivity to ethinyloestradiol (EE2) for exposure during early-life (0-48hpf) and illustrated that exposure to oestrogen (EE2) during early life (0-48hpf) can enhance responsiveness (sensitivity) to different environmental oestrogens (EE2, genistein and bisphenol A) for subsequent exposures during development. These findings illustrate the importance of oestrogen exposure history in effects assessments and they have wider implications for the possible adverse effects associated with oestrogen exposure.

Chronic liver diseases, which includes alcoholic liver disease (ALD), are consistently among the top 15 leading causes of death in the United States. ALD is characterized by progression from a normal liver to fatty liver disease (hepatic steatosis), which can lead to cirrhosis, alcoholic hepatitis, and liver failure. We have identified a novel role of phosphohistidine signaling, mediated through phosphohistidine phosphatase 1 (PHPT1), in the onset of hepatic steatosis. We have identified PHPT1 as a target of selective oxidation following acute ethanol exposure as well as being downregulated following chronic ethanol exposure. We mapped the oxidative modification site and developed a mass-spectrometry based phosphohistidine phosphatase assay to determine the impact of PHPT1 oxidative modification during acute ethanol exposure. To further understand the role of PHPT1 and phosphohistidine signaling during chronic ethanol exposure, we have developed PHPT1 overexpression and knockout mouse models. These mouse models were characterized using mass spectrometry-based proteomics. They were then utilized in a 10-day chronic ethanol plus binge model to determine the impact of PHPT1 expression on the onset of ethanol-induced hepatic steatosis. In addition, advanced mass spectrometry-based phenotypic characterization was performed on the treated liver tissues to determine the key regulators and canonical pathways influencing phosphohistidine signaling during chronic ethanol exposure. We have evidence to suggest that PHPT1 overexpression plays a protective role in the onset of hepatic steatosis, the PHPT1 heterozygous model is more susceptible to liver damage, and the complete knockout model is embryonically lethal. Additionally, we have identified novel pathways and regulators involved in phosphohistidine signaling during the development of ethanol-induced hepatic steatosis.

Space missions increasingly require sophisticated guidance, navigation and control algorithms, the development of which is reliant on verification and validation (V&V) techniques to ensure mission safety and success. A crucial element of V&V is the assessment of control system robust performance in the presence of uncertainty. In addition to estimating average performance under uncertainty, it is critical to determine the worst case performance. Industrial V&V approaches typically employ mu-analysis in the early control design stages, and Monte Carlo simulations on high-fidelity full engineering simulators at advanced stages of the design cycle. While highly capable, such techniques present a critical gap between pessimistic worst case estimates found using analytical methods, and the optimistic outlook often presented by Monte Carlo runs. Conservative worst case estimates are problematic because they can demand a controller redesign procedure, which is not justified if the poor performance is unlikely to occur. Gaining insight into the probability associated with the worst case performance is valuable in bridging this gap. It should be noted that due to the complexity of industrial-scale systems, V&V techniques are required to be capable of efficiently analysing non-linear models in the presence of significant uncertainty. As well, they must be computationally tractable. It is desirable that such techniques demand little engineering effort before each analysis, to be applied widely in industrial systems. Motivated by these factors, this thesis proposes and develops an efficient algorithm, based on the cross entropy simulation method. The proposed algorithm efficiently estimates the probabilities associated with various performance levels, from nominal performance up to degraded performance values, resulting in a curve of probabilities associated with various performance values. Such a curve is termed the probability profile of performance (PPoP), and is introduced as a tool that offers insight into a control system's performance, principally the probability associated with the worst case performance. The cross entropy-based robust performance analysis is implemented here on various industrial systems in European Space Agency-funded research projects. The implementation on autonomous rendezvous and docking models for the Mars Sample Return mission constitutes the core of the thesis. The proposed technique is implemented on high-fidelity models of the Vega launcher, as well as on a generic long coasting launcher upper stage. In summary, this thesis (a) develops an algorithm based on the cross entropy simulation method to estimate the probability associated with the worst case, (b) proposes the cross entropy-based PPoP tool to gain insight into system performance, (c) presents results of the robust performance analysis of three space industry systems using the proposed technique in conjunction with existing methods, and (d) proposes an integrated template for conducting robust performance analysis of linearised aerospace systems.