Mammary tumors are the most common neoplasms in female dogs. Until now, histological diagnosis has been performed using an optical microscope and human assessment. This study aims to explore the effectiveness of Artificial Intelligence (AI) in the histological diagnosis of canine mammary tumors. Particularly, applying neural networks, the AI was trained with 23 different histological subtypes of mammary lesions and tested on portions derived from the same cases. The results showed a diagnostic accuracy ranging from 95.7% to 100%, with the latter value achieved for 16 types of lesions. Furthermore, through a survey conducted with 12 veterinary pathologists, which involved the analysis of 92 images derived from the 23 histological subtypes, it was possible to compare the diagnostic capabilities of the human eye with those of the AI.
I tumori mammari rappresentano la neoplasia più comune nelle cagne. Fino ad oggi, la diagnosi istologica è stata effettuata tramite microscopio ottico e valutazione visiva umana. Questo studio intende esplorare l'efficacia dell'Intelligenza Artificiale (AI) nella diagnosi istologica dei tumori mammari canini. Nello specifico, utilizzando reti neurali, l’AI è stata addestrata con 23 sottotipi istologici di lesioni mammarie differenti e quindi testata su porzioni delle medesime. I risultati hanno mostrato un'accuratezza diagnostica compresa tra il 95,7% e il 100%, con quest’ultimo valore raggiunto per 16 lesioni. Inoltre, tramite un sondaggio condotto su 12 patologi veterinari, che ha previsto l'analisi di 92 immagini derivanti dai 23 sottotipi istologici menzionati, è stato possibile confrontare le capacità diagnostiche dell'occhio umano con quelle dell'intelligenza artificiale.
Tumore mammario canino: diagnosi automatizzata con l'ausilio di reti neurali su vetrini istologici digitalizzati
BERTONAZZI, BEATRICE
2023/2024
Abstract
Mammary tumors are the most common neoplasms in female dogs. Until now, histological diagnosis has been performed using an optical microscope and human assessment. This study aims to explore the effectiveness of Artificial Intelligence (AI) in the histological diagnosis of canine mammary tumors. Particularly, applying neural networks, the AI was trained with 23 different histological subtypes of mammary lesions and tested on portions derived from the same cases. The results showed a diagnostic accuracy ranging from 95.7% to 100%, with the latter value achieved for 16 types of lesions. Furthermore, through a survey conducted with 12 veterinary pathologists, which involved the analysis of 92 images derived from the 23 histological subtypes, it was possible to compare the diagnostic capabilities of the human eye with those of the AI.File | Dimensione | Formato | |
---|---|---|---|
Tesi_Bertonazzi .pdf
accesso riservato
Dimensione
8.22 MB
Formato
Adobe PDF
|
8.22 MB | Adobe PDF |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/70931