The application of computer-aided detection algorithms to HRCT (High Resolution Computed Tomography) lung imaging has the potential to objectively and quantitatively detect pulmonary pathology. We aim to evaluate the correlation between the extent of disease patterns quantified via a texture-based quantification method and pulmonary function tests (PFTs). 27 chronic hypersensitivity pneumonitis (cHP) patients (FEV1% predicted=62±16, FVC% predicted=57±17, TLC% predicted=65±17) were scanned via HRCT at full-inspiration. Regions of interest (ROIs) were marked as normal (NOR), ground glass opacity (GGO), reticulation (RET), consolidation (CON), honeycombing (HB) and air trapping (AT) by an expert pneumologist. For each ROI, statistical, morphological and fractal parameters were computed to train and test a Bayesian classifier. Then, the classifier was used to quantify the extent of each class in each patient, that was correlated via Spearman correlation analysis to PFTs. Classifier sensitivity was: AT 93.5±7.6%, CON 100±0%, GGO 95.7±6.1%, HB 94.6±5.9%, RET 92.6±7.9 and NOR 76.2±12.4; precision was: AT 86.1±7.9%, CON 99.5±2.2%, GGO 89.4±8%, HB 94.2±7.1%, RET 98.4±3.8% and NOR 89±9.5%. In the overall group, NOR, GGO and HB correlated to FVC% predicted (respectively r=0.576, p=0.002; r=-0.538, p=0.004; r=-0.480, p=0.01). NOR, GGO and HB correlated to TLC% predicted (respectively r=0.522, p=0.005; r=-0.505, p=0.007; r=-0.406, p=0.036). Texture analysis can differentiate pathological classes in cHP and the extent of some classes correlates to PFTs. Texture analysis may be used as a quantitative diagnostic tool in cHP.
L’applicazione degli algoritmi di rilevamento delle patologie nell’ambito dell’HRCT (High Resolution Computed Tomography) polmonare ha la potenzialità di quantificare oggettivamente le diverse patologie polmonari. Il nostro scopo è di valutare la correlazione tra l’estensione dei pattern patologici quantificati attraverso un metodo basato sulla texture analysis e i test di funzionalità polmonare (PFTs). 27 pazienti diagnosticati con polmonite da ipersensibilità cronica (cHP) (FEV1% predetto=62±16, FVC% predetto=57±17, TLC% predetto=65±17) sono stati acquisiti tramite HRCT a massima inspirazione. Le regioni di interesse (ROI) sono state etichettate come normale (NOR), opacità di ground glass (GGO), reticolazione (RET), consolidamento (CON), honeycombing (HB) e cisti d’aria (AT) da un esperto pneumologo. Per ogni ROI, i parametri statistici, morfologici e frattali sono stati calcolati per addestrare e testare un classificatore Bayesiano. Successivamente, il classificatore è stato usato per quantificare l’estensione di ogni classe all’interno di ogni paziente, e i risultati ottenuti sono stati correlati attraverso l’analisi di correlazione di Spearman con i PFTs. La sensitività del classificatore è risultata essere: AT 93.5±7.6%, CON 100±0%, GGO 95.7±6.1%, HB 94.6±5.9%, RET 92.6±7.9 e NOR 76.2±12.4; mentre la precisione: AT 86.1±7.9%, CON 99.5±2.2%, GGO 89.4±8%, HB 94.2±7.1%, RET 98.4±3.8% e NOR 89±9.5%. Tra tutte le classi, NOR, GGO e HB risultano essere correlate con FVC% predetto (rispettivamente r=0.576, p=0.002; r=-0.538, p=0.004; r=-0.480, p=0.01). NOR, GGO e HB risultano essere correlate con TLC% predetto (rispettivamente r=0.522, p=0.005; r=-0.505, p=0.007; r=-0.406, p=0.036). La texture analysis può differenziare le classi patologiche nel cHP e l’estensione di alcune classi correla con i PFTs. La texture analysis può essere usata come metodo diagnostico quantitativo nel cHP.
Texture analysis for quantitative CT in hypersensitivity pneumonitis
ANTONIAZZA, ALESSIO;BERETTA, DAVIDE
2017/2018
Abstract
The application of computer-aided detection algorithms to HRCT (High Resolution Computed Tomography) lung imaging has the potential to objectively and quantitatively detect pulmonary pathology. We aim to evaluate the correlation between the extent of disease patterns quantified via a texture-based quantification method and pulmonary function tests (PFTs). 27 chronic hypersensitivity pneumonitis (cHP) patients (FEV1% predicted=62±16, FVC% predicted=57±17, TLC% predicted=65±17) were scanned via HRCT at full-inspiration. Regions of interest (ROIs) were marked as normal (NOR), ground glass opacity (GGO), reticulation (RET), consolidation (CON), honeycombing (HB) and air trapping (AT) by an expert pneumologist. For each ROI, statistical, morphological and fractal parameters were computed to train and test a Bayesian classifier. Then, the classifier was used to quantify the extent of each class in each patient, that was correlated via Spearman correlation analysis to PFTs. Classifier sensitivity was: AT 93.5±7.6%, CON 100±0%, GGO 95.7±6.1%, HB 94.6±5.9%, RET 92.6±7.9 and NOR 76.2±12.4; precision was: AT 86.1±7.9%, CON 99.5±2.2%, GGO 89.4±8%, HB 94.2±7.1%, RET 98.4±3.8% and NOR 89±9.5%. In the overall group, NOR, GGO and HB correlated to FVC% predicted (respectively r=0.576, p=0.002; r=-0.538, p=0.004; r=-0.480, p=0.01). NOR, GGO and HB correlated to TLC% predicted (respectively r=0.522, p=0.005; r=-0.505, p=0.007; r=-0.406, p=0.036). Texture analysis can differentiate pathological classes in cHP and the extent of some classes correlates to PFTs. Texture analysis may be used as a quantitative diagnostic tool in cHP.| File | Dimensione | Formato | |
|---|---|---|---|
|
2019_04_Antoniazza_Beretta.pdf
accessibile in internet per tutti
Descrizione: Thesis text
Dimensione
6.64 MB
Formato
Adobe PDF
|
6.64 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/146205