Virtual Library

Start Your Search

Pretesh R Patel



Author of

  • +

    P2.03 - Biology (Not CME Accredited Session) (ID 952)

    • Event: WCLC 2018
    • Type: Poster Viewing in the Exhibit Hall
    • Track:
    • Presentations: 1
    • Moderators:
    • Coordinates: 9/25/2018, 16:45 - 18:00, Exhibit Hall
    • +

      P2.03-07 - Radiomic Signatures Linked to Genetic Alterations as Detected by Next-Generation Sequencing: A Radiogenomics Analysis of Early-Stage NSCLC (ID 11858)

      16:45 - 18:00  |  Author(s): Pretesh R Patel

      • Abstract
      • Slides

      Background

      Radiomics uses large-scale quantitative analysis of extracted image-based features to identify tumor phenotypes. Such frameworks based on computed tomography (CT) have identified informative features in lung cancer related to treatment response and prognosis. Molecular profiles have been increasingly recognized in non-small cell lung carcinoma (NSCLC) as predictors of clinical outcomes. The aim of this study is to isolate radiomic signatures which are associated with genomic alterations discoverable by extended panel testing.

      a9ded1e5ce5d75814730bb4caaf49419 Method

      Patients with early-stage NSCLC who received definitive local treatment and underwent next-generation sequencing were included for this analysis. 35 genomic alterations representing 32 genes were evaluated using SNaPshot (Life Technologies), TruSight (Illumina Inc), Guardant360 (Guardant Health) or a FISH panel of ALK, MET, ROS1 and RET. Regions of interest including each primary tumor volume were delineated on diagnostic CTs datasets; they were defined as either tumor only (TO), or tumor with a 1cm anatomically-modified anisotropic margin (TM). Measures describing delineation and tumor appearance within each volume were extracted using in-house software, recording HU characteristics, inhomogeneity, and first and second-degree texture statistics. 91 structural, textural, and intensity features were extracted. Univariate logistic regression was performed to test the performance between each extracted feature and the presence of genomic alterations.

      4c3880bb027f159e801041b1021e88e8 Result

      40 patients with diagnostic CTs available for feature extraction were included for analysis. The rate of mutation prevalence in the most commonly altered genes were: TP53 (52.1%), cMET (34.5%), KRAS (27%), PTEN (13.8%), EGFR (13.2%), and MET (12.1%). 5 features were associated with TP53 mutations using TO-extracted features, and 4 using the TM-based approach. cMET was linked to Hounsfield minimum (odds ratio [OR] 1.01, p=0.027), and Sobel minimum (p=0.049) using the TO approach. Long run emphasis standard deviation (STD) [OR 0.36, p=0.04], and long run low grey level emphasis STD (OR 0.37, p=0.04) correlated with cMET mutations using the TM approach. Laplacian sharpening mean was associated with the presence of EGFR mutations using both extraction approaches; finding OR 1.10 (p=0.04, TO method) and OR 1.12 (p=0.02, TM method). No other features were linked to EGFR alterations. Gradient magnitude variance was associated with PTEN using tumor-with-margin volumes (OR 0.83, p=0.028). No features were significantly linked to KRAS mutations using either extraction approach.

      8eea62084ca7e541d918e823422bd82e Conclusion

      Multiple radiomic features were associated with TP53, cMET, EGFR, and PTEN mutations. Integration of such signatures may help inform prognosis for a heterogenous cohort of patients with early-stage non-small cell lung carcinoma.

      6f8b794f3246b0c1e1780bb4d4d5dc53

      Only Active Members that have purchased this event or have registered via an access code will be able to view this content. To view this presentation, please login or select "Add to Cart" and proceed to checkout.