Presentation Authors: Vinay Duddalwar*, Los Angeles, CA, Seth Lerner, Houston, TX, Erich Huang, Bethesda, MD, Bino Varghese, Kevin King, Steven Cen, Darryl Hwang, Los Angeles, CA, Ersan Altun, Chapel Hill, NC, Tharakeswara Bathala, Houston, TX, Steven Kennish, Sheffield, United Kingdom, Juan Ibarra Rovira, Houston, TX, Fabiano Lucchesi, Barretos, Brazil, Valdair Francisco Muglia, Sao Paulo, Brazil, Stephen Thomas, Chicago, IL, Raghu Vikram, Houston, TX, Hebert Alberto Vargas, New York,, NY, Brenda Fevrier-Sullivan, Justin Kirby, Carl Jaffe, John Freymann, Bethesda, MD
Introduction: Medical imaging is routinely used for the diagnosis and treatment of cancer. Genomic testing has led to hallmark mutations being identified to characterize certain cancers including muscle-invasive bladder cancer (MIBC). However, owing to the rapidly evolving mutational landscape, it is currently impossible to obtain a reliable assessment of tumor heterogeneity. Imaging is a clinically accepted technique in both clinical trials and routine practice and, easily incorporated in longitudinal studies. Establishing reliable associations between imaging-based metrics (radiomics) and frequently mutated cancer genes can be of great value in cancer management, particularly for MIBC.
Methods: We integrated genomics data from The Cancer Genome Atlas (TCGA) with CT data from The Cancer Imaging Archive (TCIA) in 89 biopsy-proven MIBC cancer patients to find associations between hallmark MIBC mutations and radiomic metrics. Using an in-house developed Matlab code, a CT-based radiomics panel comprising of texture metrics extracted using six different methods: histogram analysis, 2D- Gray-level co-occurrence matrix (GLCM), Gray-level difference matrix (GLDM) and Gray-level run-length matrix (GLRLM) and 2D- Fast Fourier Transform (FFT) analyses was applied to the segmented images. The extracted 488 texture features were associated with the genetic profiles of 16 mutations of MIBC namely ARID1A, ASXL2, ATM, CASP8, CDKN1A, CREBBP, ELF3, ERBB2, FGFR3, KDM6A, KMT2D, PIK3CA, RB1, RHOB, TP53 and EP300 which were found in atleast 10% of cases in this cohort. 488 x 16 pairs of t-tests or Wilcoxon rank sum tests were used to create a panel of associations and studied using a heatmap of univariate p-values.
Results: Amongst the 16 mutations, radiomic metrics showed strong associations with ARIDIA, FGFR3 and EP300 mutations, evident from the increased percent of radiomic metrics with a p-value < 0.05 i.e., 16.6%, 14.6% and 16.6% respectively. Of all the radiomics metrics, GLDM2D showed most associations particularly with ARIDIA, FGFR3 and EP300 mutations.
Conclusions: Our discovery analysis study reveals promising findings with significant associations between non-invasive quantitative imaging metrics and genetic mutations. This highlights the need for further studies investigating this association.