Background:According to published data, radiomics features differ between lesionsof refractory/relapsing HL patients from those of long-term responders. However,several methodological aspects have not been elucidated yet.Purpose:The study aimed at setting up a methodological framework in radiomicsapplications in Hodgkin’s lymphoma (HL), especially at (a) developing a novel featureselection approach, (b) evaluating radiomic intra-patient lesions’similarity, and (c)classifying relapsing refractory (R/R) vs non-(R/R) patients.Methods:We retrospectively included 85 patients (male:female = 52:33; median age35 years, range 19–74). LIFEx (www.lifexsoft.org) was used for [18F]FDG-PET/CTsegmentation and feature extraction. Features werea-prioriselected if they werehighly correlated or uncorrelated to the volume. Principal component analysis-transformed features were used to build the fingerprints that were tested to assesslesions’similarity, using thesilhouette. For intra-patient similarity analysis, we usedpatients having multiple lesions only. To classify patients as non-R/R and R/R, thefingerprint considering one single lesion (fingerprint_One) and all lesions(fingerprint_All) was tested using Random Undersampling Boosting of Tree Ensemble(RUBTE).Results:HL fingerprints included up to 15 features. Intra-patient lesion similarityanalysis resulted in mean/median silhouette values below 0.5 (low similarityespecially in the non-R/R group). In the test set, the fingerprint_One classificationaccuracy was 62% (78% sensitivity and 53% specificity); the classification by RUBTEusing fingerprint_All resulted in 82% accuracy (70% sensitivity and 88% specificity).Conclusions:Lesion similarity analysis was developed, and it allowed to demonstratethat HL lesions were not homogeneous within patients in terms of radiomics signature.Therefore, a random target lesion selection should not be adopted for radiomicsapplications. Moreover, the classifier to predict R/R vs non-R/R performed the bestwhen all the lesions were used.

Methodological framework for radiomics applications in Hodgkin’s lymphoma / M. Sollini, M. Kirienko, L. Cavinato, F. Ricci, M. Biroli, F. Ieva, L. Calderoni, E. Tabacchi, C. Nanni, P. Luigi Zinzani, S. Fanti, A. Guidetti, A. Alessi, P. Corradini, E. Seregni, C. Carlo Stella, A. Chiti. - In: EUROPEAN JOURNAL OF HYBRID IMAGING. - ISSN 2510-3636. - 4:1(2020).

Methodological framework for radiomics applications in Hodgkin’s lymphoma

F. Ieva;A. Guidetti;P. Corradini;C. Carlo Stella;
2020

Abstract

Background:According to published data, radiomics features differ between lesionsof refractory/relapsing HL patients from those of long-term responders. However,several methodological aspects have not been elucidated yet.Purpose:The study aimed at setting up a methodological framework in radiomicsapplications in Hodgkin’s lymphoma (HL), especially at (a) developing a novel featureselection approach, (b) evaluating radiomic intra-patient lesions’similarity, and (c)classifying relapsing refractory (R/R) vs non-(R/R) patients.Methods:We retrospectively included 85 patients (male:female = 52:33; median age35 years, range 19–74). LIFEx (www.lifexsoft.org) was used for [18F]FDG-PET/CTsegmentation and feature extraction. Features werea-prioriselected if they werehighly correlated or uncorrelated to the volume. Principal component analysis-transformed features were used to build the fingerprints that were tested to assesslesions’similarity, using thesilhouette. For intra-patient similarity analysis, we usedpatients having multiple lesions only. To classify patients as non-R/R and R/R, thefingerprint considering one single lesion (fingerprint_One) and all lesions(fingerprint_All) was tested using Random Undersampling Boosting of Tree Ensemble(RUBTE).Results:HL fingerprints included up to 15 features. Intra-patient lesion similarityanalysis resulted in mean/median silhouette values below 0.5 (low similarityespecially in the non-R/R group). In the test set, the fingerprint_One classificationaccuracy was 62% (78% sensitivity and 53% specificity); the classification by RUBTEusing fingerprint_All resulted in 82% accuracy (70% sensitivity and 88% specificity).Conclusions:Lesion similarity analysis was developed, and it allowed to demonstratethat HL lesions were not homogeneous within patients in terms of radiomics signature.Therefore, a random target lesion selection should not be adopted for radiomicsapplications. Moreover, the classifier to predict R/R vs non-R/R performed the bestwhen all the lesions were used.
Lymphoma, PET/CT, Radiomics, Similarity, Feature selection, Silhouette,Response prediction, Outcome prediction
Settore MED/15 - Malattie del Sangue
Settore MED/36 - Diagnostica per Immagini e Radioterapia
Article (author)
File in questo prodotto:
File Dimensione Formato  
Sollini_et_al-2020-European_Journal_of_Hybrid_Imaging.pdf

accesso aperto

Tipologia: Publisher's version/PDF
Dimensione 1.42 MB
Formato Adobe PDF
1.42 MB 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/2434/755186
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 7
  • ???jsp.display-item.citation.isi??? 5
social impact