The islets of Langerhans regulate glucose homeostasis through insulin and glucagon secretion. Understanding the metabolic dynamics of α and β cells is crucial for diabetes research. This work presents two complementary studies using fluorescence lifetime imaging microscopy (FLIM) and machine learning to analyze metabolic responses in living human islets. In the first study, two-photon FLIM was used to measure NAD(P)H autofluorescence before and after glucose stimulation. By integrating phasor-FLIM with immunofluorescence-based cell-type identification, metabolic shifts in 312 α and 654 β cells were quantified. β cells generally exhibited an increase in enzyme-bound NAD(P)H, suggesting enhanced oxidative phosphorylation, while cells metabolic shifts revealed to be dependent on donor's insulin secretion efficiency, indicating a link between single-cell metabolism and whole-islet function. The second study introduces a machine-learning model for identifying α and β cells in label-free infrared micrographs of living islets. A boosted decision-tree model (XGBoost) achieved a ROC_AUC score of 0.86, with high β-cell recognition (94%) and benchmark-matching α-cell performance (75%). Collectively, these findings provide a comprehensive tool for investigation of human Langerhans islets and demonstrate the potential of FLIM and machine learning to advance islet cell metabolism research, offering new insights into diabetes diagnostics and therapy.
Le isole di Langerhans regolano l’omeostasi del glucosio attraverso la secrezione di insulina e glucagone. Comprendere le dinamiche metaboliche delle cellule α e β è cruciale per la ricerca sul diabete. Questo lavoro presenta due studi complementari che utilizzano la microscopia a tempo di vita della fluorescenza (FLIM) e il machine learning per analizzare la risposta metaboliche nelle isole pancreatiche umane. Nel primo studio, la FLIM a due fotoni è stata utilizzata per misurare l'autofluorescenza del NAD(P)H prima e dopo la stimolazione con glucosio. Integrando l'analisi phasor-FLIM con l’identificazione della tipologia cellulare tramite immunofluorescenza, sono stati quantificati i cambiamenti metabolici in 312 cellule α e 654 cellule β. Le cellule β hanno generalmente mostrato un aumento del NAD(P)H legato a enzimi, suggerendo un incremento della fosforilazione ossidativa, mentre il comportamento delle α cellule si è rivelato dipendente della capacità di secrezione di insulina del donatore, indicando un legame tra il metabolismo delle singole cellule e la funzione dell’intera isola. Il secondo studio introduce un modello di machine learning per identificare le cellule α e β in immagini label-free di isole pancreatiche viventi. Un modello ad albero decisionale potenziato (XGBoost) ha raggiunto un punteggio ROC_AUC di 0.86, rivelando un'elevata precisione per il riconoscimento delle cellule β (94%) e una prestazione pari al benchmark per le cellule α (75%). Nel complesso, questi risultati forniscono un protocollo per lo studio delle isole di Langerhans umane e dimostrano il potenziale della FLIM e del machine learning nell'avanzamento della ricerca sul metabolismo delle cellule delle isole, offrendo nuove prospettive per la diagnosi e la terapia del diabete.
Label-free optical microscopy of alpha and beta cells in human Langerhans islets
AZZARELLO, Fabio
2025
Abstract
The islets of Langerhans regulate glucose homeostasis through insulin and glucagon secretion. Understanding the metabolic dynamics of α and β cells is crucial for diabetes research. This work presents two complementary studies using fluorescence lifetime imaging microscopy (FLIM) and machine learning to analyze metabolic responses in living human islets. In the first study, two-photon FLIM was used to measure NAD(P)H autofluorescence before and after glucose stimulation. By integrating phasor-FLIM with immunofluorescence-based cell-type identification, metabolic shifts in 312 α and 654 β cells were quantified. β cells generally exhibited an increase in enzyme-bound NAD(P)H, suggesting enhanced oxidative phosphorylation, while cells metabolic shifts revealed to be dependent on donor's insulin secretion efficiency, indicating a link between single-cell metabolism and whole-islet function. The second study introduces a machine-learning model for identifying α and β cells in label-free infrared micrographs of living islets. A boosted decision-tree model (XGBoost) achieved a ROC_AUC score of 0.86, with high β-cell recognition (94%) and benchmark-matching α-cell performance (75%). Collectively, these findings provide a comprehensive tool for investigation of human Langerhans islets and demonstrate the potential of FLIM and machine learning to advance islet cell metabolism research, offering new insights into diabetes diagnostics and therapy.| File | Dimensione | Formato | |
|---|---|---|---|
|
Tesi.pdf
accesso aperto
Licenza:
Tutti i diritti riservati
Dimensione
13.28 MB
Formato
Adobe PDF
|
13.28 MB | Adobe PDF | Visualizza/Apri |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14242/304295
URN:NBN:IT:SNS-304295