Background. In order to help physicians and radiologists in diagnosing pneumonia, deep learning and other artificial intelligence methods have been described in several researches to solve this task. The main objective of the present study is to build a stacked hierarchical model by combining several models in order to increase the procedure accuracy. Methods. Firstly, the best convolutional network in terms of accuracy were evaluated and described. Later, a stacked hierarchical model was built by using the most relevant features extracted by the selected two models. Finally, over the stacked model with the best accuracy, a hierarchically dependent second stage model for inner-classification was built in order to detect both inflammation of the pulmonary alveolar space (lobar pneumonia) and interstitial tissue involvement (interstitial pneumonia). Results. The study shows how the adopted staked model lead to a higher accuracy. Having a high accuracy on pneumonia detection and classification can be a paramount asset to treat patients in real health-care environments. Conclusions. Despite some limits, our findings support the notion that deep learning methods can be used to simplify the diagnostic process and improve disease management.

Hierarchical convolutional models for automatic pneumonia diagnosis based on X-ray images: new strategies in public health / Maselli, Gianluca; Bertamino, Enrico; Capalbo, Carlo; Mancini, Rita; Orsi, Giovanni Battista; Napoli, Christian; Napoli, Christian. - In: ANNALI DI IGIENE MEDICINA PREVENTIVA E DI COMUNITÀ. - ISSN 1120-9135. - 33:6(2021), pp. 644-655. [10.7416/ai.2021.2467]

Hierarchical convolutional models for automatic pneumonia diagnosis based on X-ray images: new strategies in public health

Enrico Bertamino
Secondo
Conceptualization
;
Carlo Capalbo
Membro del Collaboration Group
;
Rita Mancini
Membro del Collaboration Group
;
Giovanni Battista Orsi
Membro del Collaboration Group
;
Christian Napoli
Penultimo
Supervision
;
Christian Napoli
Ultimo
Supervision
2021

Abstract

Background. In order to help physicians and radiologists in diagnosing pneumonia, deep learning and other artificial intelligence methods have been described in several researches to solve this task. The main objective of the present study is to build a stacked hierarchical model by combining several models in order to increase the procedure accuracy. Methods. Firstly, the best convolutional network in terms of accuracy were evaluated and described. Later, a stacked hierarchical model was built by using the most relevant features extracted by the selected two models. Finally, over the stacked model with the best accuracy, a hierarchically dependent second stage model for inner-classification was built in order to detect both inflammation of the pulmonary alveolar space (lobar pneumonia) and interstitial tissue involvement (interstitial pneumonia). Results. The study shows how the adopted staked model lead to a higher accuracy. Having a high accuracy on pneumonia detection and classification can be a paramount asset to treat patients in real health-care environments. Conclusions. Despite some limits, our findings support the notion that deep learning methods can be used to simplify the diagnostic process and improve disease management.
2021
accuracy; convolutional neural network; diagnostic process
01 Pubblicazione su rivista::01a Articolo in rivista
Hierarchical convolutional models for automatic pneumonia diagnosis based on X-ray images: new strategies in public health / Maselli, Gianluca; Bertamino, Enrico; Capalbo, Carlo; Mancini, Rita; Orsi, Giovanni Battista; Napoli, Christian; Napoli, Christian. - In: ANNALI DI IGIENE MEDICINA PREVENTIVA E DI COMUNITÀ. - ISSN 1120-9135. - 33:6(2021), pp. 644-655. [10.7416/ai.2021.2467]
File allegati a questo prodotto
File Dimensione Formato  
Maselli_Hierarchical_2021.pdf

solo gestori archivio

Tipologia: Versione editoriale (versione pubblicata con il layout dell'editore)
Licenza: Tutti i diritti riservati (All rights reserved)
Dimensione 491.25 kB
Formato Adobe PDF
491.25 kB Adobe PDF   Contatta l'autore

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11573/1565103
Citazioni
  • ???jsp.display-item.citation.pmc??? 1
  • Scopus 5
  • ???jsp.display-item.citation.isi??? 2
social impact