Virtual Library

Start Your Search

H. Beaumont



Author of

  • +

    MA01 - Improvement and Implementation of Lung Cancer Screening (ID 368)

    • Event: WCLC 2016
    • Type: Mini Oral Session
    • Track: Radiology/Staging/Screening
    • Presentations: 1
    • +

      MA01.05 - Predictive Performances of NELSON Screening Program Based on Clinical, Metrological and Population Statistics (ID 4688)

      11:00 - 12:30  |  Author(s): H. Beaumont

      • Abstract
      • Presentation
      • Slides

      Background:
      The balance of benefits and harms of screening programs depends on multiple factors such as the scenario of patient selection, the triage algorithm and the imaging methods. Because of the multifactorial nature of the outcome of screening programs, it is important to evaluate the performance of its components. We modeled the triage algorithm of the NELSON program for lung cancer screening in different scenarios in order to assess the robustness of the chosen approach. We are looking to develop a model that allows for testing the imaging protocol performance using various high-risk screening populations. Our Objective is to work out a simulator adaptive to multiple screening scenarios. In a first step, we tested a simulation of the NELSON triage algorithm by using published statistics as input data: the distribution of nodule size, the precision of nodule volume measurements and the distribution of nodules growth.

      Methods:
      We modeled the baseline round of NELSON triage algorithm. We simulated 10,000,000 ground truth (GT) data where the axial diameter of nodules followed a chi2 (df=1) distribution between 3 mm and 20 mm. For each of the GT nodule, we modeled also a chi2 (df=1) distribution of volume doubling time between 90 and 1000 days. We included into the model a Gaussian distribution of the time between visits (average: 105 days, standard deviation: 5 days). We modeled volume measurement of the nodules by adding a Gaussian random error as documented by the Quantitative Imaging Biomarker Alliance (QIBA) screening profile. We performed a by-nodule comparison between nodule classification by the triage algorithm and the corresponding GT in the first round. At each step of the triage algorithm, we evaluated Sensitivity (Se), Specificity (Sp), Positive Predictive Value (PPV) and Negative Predictive Value (NPV).

      Results:
      Sensitivity of the triage algorithm for classifying nodules into size categories was for 96,6% for NODCAT2, 86.9% for NODCAT3 and 90.7% for NODCAT4. Classification of GROWCAT C yielded Se=66.2% / Sp=21.2%. We found an overall performance of the NELSON triage algorithm of Se/Sp 94.0%/80.3%.PPV was 11.3%, and NPV was 99.8%

      Conclusion:
      Mathematical modeling gives valuable insights into the performance of different components of triage algorithms in lung cancer screening. We found a markedly different test performance for size versus growth assessment of the NELSON triage algorithm. Future work will extent the model to non-solid nodules and multiple rounds of screening. Moreover, it may have the potential to optimize triage algorithms in the design of screening programs.

      Only 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, select "Add to Cart" and proceed to checkout. If you would like to become a member of IASLC, please click here.

      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.