Discovery of pre-therapy 2-deoxy-2-18F-fluoro-D-glucose positron emission tomography-based radiomics classifiers of survival outcome in non-small-cell lung cancer patients

• DOI: 10.1007/s00259-018-4139-4
• Publication Date: 2018-09-01T03:57:49Z
• Authors: Mubarik A. Arshad; Andrew Thornton; Haonan Lu; Henry Tam; Kathryn Wallitt; Nicola Rodgers; Andrew Scarsbrook; Garry McDermott; Gary J. Cook; David Landau; Sue Chua; Richard O’Connor; Jeanette Dickson; Danielle A. Power; Tara D. Barwick; Andrea Rockall; Eric O. Aboagye
• Keywords: Adult; Male; Lung Neoplasms; Survival; PREDICTION; Image Processing; 0299 Other Physical Sciences; ED AMERICAN-COLLEGE; TEXTURAL FEATURES; NSCLC; PARAMETERS; 03 medical and health sciences; Computer-Assisted; 0302 clinical medicine; Fluorodeoxyglucose F18; Carcinoma, Non-Small-Cell Lung; Positron Emission Tomography Computed Tomography; 616; 80 and over; MANAGEMENT; Image Processing, Computer-Assisted; Humans; HETEROGENEITY; Non-Small-Cell Lung; FDG-PET; Risk stratification; Aged; Aged, 80 and over; Science & Technology; Radiomics; Carcinoma; Radiology, Nuclear Medicine & Medical Imaging; 1103 Clinical Sciences; Middle Aged; Survival Analysis; 3. Good health; VARIABILITY; Nuclear Medicine & Medical Imaging; PET; TUMOR VOLUME; Original Article; Female; Radiology; Life Sciences & Biomedicine; RADIOTHERAPY

The aim of this multi-center study was to discover and validate radiomics classifiers as image-derived biomarkers for risk stratification non-small-cell lung cancer (NSCLC). Pre-therapy PET scans from a total 358 Stage I–III NSCLC patients scheduled radiotherapy/chemo-radiotherapy acquired between October 2008 December 2013 were included in seven-institution study. A semi-automatic threshold method used segment the primary tumors. Radiomics predictive derived training set 133 using TexLAB v2. Least absolute shrinkage selection operator (LASSO) regression analysis data dimension reduction feature vector (FV) discovery. Multivariable performed establish relationship FV, stage overall survival (OS). Performance optimal FV tested an independent validation 204 patients, further 21 (TESTI) patients. Of 249 died within follow-up period [median 22 (range 0–85) months]. From each tumor, 665 three-dimensional features seven gray levels extracted. most discovered (FVX) known prognostic factors, such tumor volume, interest studies, invariant type PET/CT manufacturer. Using median cut-off, FVX predicted 14-month difference cohort (N = 204, p 0.00465; HR 1.61, 95% CI 1.16–2.24). In TESTI cohort, smaller that presented with unusually poor I cancers, correctly indicated lack 21, 0.501). contrast classifier, clinically routine variables including SUVmax, SUVmean SUVpeak lacked any information. PET-based pre-treatment imaging possess intrinsic information radiotherapy/chemo-radiotherapy.