In this paper we describe a method processing Postero Anterior chest radiographs to extract a set of nodule candidate regions characterized by a low cardinality and a high sensitivity ratio. It is based on two consecutive multiscale procedures to first enhance the visibility of the nodules and then extract a first set of candidates. To reduce its cardinality two different methods, applied to a set of most representative features, are described and compared: a rule based system and a feed-forward neural network trained by back-propagation. Both the systems had similarly good results. The method has been developed and tested on a standard database containing 154 radiographs of patients with lung nodules and 93 with no nodules. The final set of candidates detected contains about 10000 regions, approximately 40 per image, and 8 true positives out of 154 are lost; the sensitivity ratio of the system is then equal to 0.95%.

A nodule detection system for postero-anterior chest radiographs / P. Campadelli, E. Casiraghi - In: Modelling, computation and optimization in information systems and management sciences : MCO 2004 : july 1-3, 2004 : proceedings / L.T.H. An, P.D. Tao. - London : Hermes Science, 2004 Jul. - ISBN 1903398215. (( Intervento presentato al 5. convegno Modelling, Computation and Optimization in Information Systems and Management Sciences(MCO) tenutosi a Metz, France nel 2004.

A nodule detection system for postero-anterior chest radiographs

P. Campadelli;E. Casiraghi
2004

Abstract

In this paper we describe a method processing Postero Anterior chest radiographs to extract a set of nodule candidate regions characterized by a low cardinality and a high sensitivity ratio. It is based on two consecutive multiscale procedures to first enhance the visibility of the nodules and then extract a first set of candidates. To reduce its cardinality two different methods, applied to a set of most representative features, are described and compared: a rule based system and a feed-forward neural network trained by back-propagation. Both the systems had similarly good results. The method has been developed and tested on a standard database containing 154 radiographs of patients with lung nodules and 93 with no nodules. The final set of candidates detected contains about 10000 regions, approximately 40 per image, and 8 true positives out of 154 are lost; the sensitivity ratio of the system is then equal to 0.95%.
CAD systems ; Multiscale image enhancement ; Pattern classification
Settore INF/01 - Informatica
lug-2004
Book Part (author)
File in questo prodotto:
File Dimensione Formato  
MCOElenaCasiraghi.pdf

accesso riservato

Tipologia: Post-print, accepted manuscript ecc. (versione accettata dall'editore)
Dimensione 197.27 kB
Formato Adobe PDF
197.27 kB Adobe PDF   Visualizza/Apri   Richiedi una copia
Pubblicazioni consigliate

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/2434/179744
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact