According to the guidelines, patients with locally advanced colorectal cancer undergo neoadjuvant chemotherapy. However, response to therapy is reached only up to 30% of cases. Therefore, it would be important to predict response to therapy before treatment. In this study, we demonstrated that the simultaneous optimization of feature subset and classifier parameters on different imaging datasets (T2w, DWI and PET) could improve classification performance. On a dataset of 51 patients (21 responders, 30 non responders), we obtained an accuracy of 90%, 84% and 76% using three optimized SVM classifiers fed with selected features from PET, T2w and ADC images, respectively.

Radiomics to predict response to neoadjuvant chemotherapy in rectal cancer: influence of simultaneous feature selection and classifier optimization / Rosati, S; Gianfreda, Cm; Balestra, G; Giannini, V; Mazzetti, S; Regge, D. - ELETTRONICO. - (2018), pp. 65-68. (Intervento presentato al convegno 2018 IEEE Life Sciences Conference (LSC) tenutosi a Montreal, QC, Canada nel 28-30 Oct. 2018) [10.1109/LSC.2018.8572194].

Radiomics to predict response to neoadjuvant chemotherapy in rectal cancer: influence of simultaneous feature selection and classifier optimization

Rosati, S;Balestra, G;Giannini, V;
2018

Abstract

According to the guidelines, patients with locally advanced colorectal cancer undergo neoadjuvant chemotherapy. However, response to therapy is reached only up to 30% of cases. Therefore, it would be important to predict response to therapy before treatment. In this study, we demonstrated that the simultaneous optimization of feature subset and classifier parameters on different imaging datasets (T2w, DWI and PET) could improve classification performance. On a dataset of 51 patients (21 responders, 30 non responders), we obtained an accuracy of 90%, 84% and 76% using three optimized SVM classifiers fed with selected features from PET, T2w and ADC images, respectively.
File in questo prodotto:
File Dimensione Formato  
Retto_DEF.pdf

accesso aperto

Tipologia: 2. Post-print / Author's Accepted Manuscript
Licenza: PUBBLICO - Tutti i diritti riservati
Dimensione 533.39 kB
Formato Adobe PDF
533.39 kB Adobe PDF Visualizza/Apri
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/11583/2734973
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo