Virtual Library

Start Your Search

D. Manos



Author of

  • +

    MA 14 - Diagnostic Radiology, Staging and Screening for Lung Cancer I (ID 672)

    • Event: WCLC 2017
    • Type: Mini Oral
    • Track: Radiology/Staging/Screening
    • Presentations: 1
    • +

      MA 14.11 - Malignancy Risk Prediction of Pulmonary Nodule in Lung Cancer Screening – Diameter Or Volumetric Measurement  (ID 9113)

      15:45 - 17:30  |  Author(s): D. Manos

      • Abstract
      • Presentation
      • Slides

      Background:
      Nodule size is an important parameter to determine malignancy risk. Semi-automated size measurements have the potential to replace manual measurements due to their higher accuracy and reproducibility, and less inter/intra-user variation. However, controversy exists regarding the relative accuracy of 2D diameter versus 3D volumetric measurement to predict malignancy risk. The objective of this study is to compare nodule malignancy prediction models based on 2D mean diameter versus volumetric measurement, both generated by a CAD Software.

      Method:
      We analyzed baseline LDCT reconstructed using high spatial frequency algorithm from 1746 participants (47% women, 53% men, age: 62.5 ± 5.8 yrs) in the Pan-Canadian Early Detection of Lung Cancer Study (PanCan), who had ≥1 non-calcified nodules ≥3mm in diameter. CAD software (CIRRUS Lung Screening, Radboud University Medical Center, Nijmegen, the Netherlands) performed an automatic nodule segmentation, which could be optimised manually, measurement of mean diameter and volume was generated. Malignant or benign nodule status was confirmed by pathology or prolonged follow-up (median follow-up 5.5 years). Logistic regression models predicting cancer were prepared with one including mean diameter and the other including volume. The discrimination, the ability to classify cancer versus benign nodules correctly, was evaluated by the area under the receiver operator characteristic cure (AUC). The calibration - do predicted probabilities match observed probabilities, was assessed using Spiegelhalter’s z-test and graphically by plotting the observed and predicted mean probabilities of cancer by deciles of model risk.

      Result:
      There were in total 5878 nodules, including 119 cancers in 115 individuals. Both models gave similar predictive performances. AUC was 0.947 (95% CI 0.922-0.964) in the mean diameter model and 0.946 (95% CI 0.921-0.966) in the volumetric model (p=0.83). The calibrations were similar between the two models (figure). Figure 1



      Conclusion:
      The predictive performances of nodule malignancy prediction models using mean 2D nodule diameter and 3D volumetric data were indistinguishable.

      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.

  • +

    OA 15 - Diagnostic Radiology, Staging and Screening for Lung Cancer II (ID 684)

    • Event: WCLC 2017
    • Type: Oral
    • Track: Radiology/Staging/Screening
    • Presentations: 1
    • +

      OA 15.01 - Lung Cancer Screening: Participant Selection by Risk Model – the Pan-Canadian Study (ID 8466)

      14:30 - 16:15  |  Author(s): D. Manos

      • Abstract
      • Presentation
      • Slides

      Background:
      Retrospective studies indicate that selecting individuals for low dose computed tomography (LDCT) lung cancer screening based on a highly predictive risk model is superior to applying National Lung Screening Trial (NLST)-like criteria, which use only categorized age, pack-year and smoking quit-time information. The Pan-Canadian Early Detection of Lung Cancer Study (PanCan Study) was designed to prospectively evaluate whether individuals at high risk for lung cancer could be identified for screening using a risk prediction model. This paper describes the study design and results.

      Method:
      2537 individuals were recruited through 8 centers across Canada based on a ≥2% of lung cancer risk estimated by the PanCan model, a precursor to the validated PLCOm2012 model. Individuals were screened at baseline and 1 and 4 years post-baseline.

      Result:
      At a median 5.5 years of follow-up, 164 individuals (6.5%) were diagnosed with 172 lung cancers. This was a significantly greater percentage of persons diagnosed with lung cancers than was observed in the NLST(4.0%)(p<0·001). Compared to 57% observed in the NLST, 77% of lung cancers in the PanCan Study were early stage (I or II) (p<0.001) and to 25% in a comparable population, age 50-75 during 2007-2009 in Ontario, Canada’s largest province, (p<0·001).

      Conclusion:
      Enrolling high-risk individuals into a LDCT screening study or program using a highly predictive risk model, is efficient in identifying individuals who will be diagnosed with lung cancer and is compatible with a strong stage shift – identifying a high proportion at early, potentially curable stage. Funding This study was funded by the Terry Fox Research Institute and Canadian Partnership Against Cancer. ClinicalTrials.gov number, NCT00751660

      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.