Virtual Library

Start Your Search

N. Peled



Author of

  • +

    P1.20 - Poster Session 1 - Early Detection and Screening (ID 172)

    • Event: WCLC 2013
    • Type: Poster Session
    • Track: Imaging, Staging & Screening
    • Presentations: 1
    • +

      P1.20-006 - Non-invasive Detection of Lung Cancer from Cells in Sputum Using Cell-CT™ (ID 2624)

      09:30 - 16:30  |  Author(s): N. Peled

      • Abstract

      Background
      Cell-CT is a fully-automated, single-cell analysis system that has been developed to enable non-invasive screening for lung cancer. We report the results of a pilot clinical study using Cell-CT for lung cancer detection from patient sputum, including the determination of criteria that define specimen adequacy, the detection of abnormal cells in adequate specimens (sensitivity), and the detection of normal cells (specificity), performed automatically or further enhanced with human evaluation.

      Methods
      Twenty-seven patients with biopsy-confirmed lung cancer produced spontaneous-cough sputa (three morning-pooled) that were fixed, stained with hematoxylin following mucus dissolution, then enriched for epithelial cells using the method of fluorescence-activated cell sorting. The specimens were analyzed using Cell-CT, which computes 3D digital images of single cells through tomographic reconstruction with isometric, sub-micron resolution (see figure). The 3D cell images were automatically interrogated to measure morphometric biosignatures for lung cancer. 3D feature measurements were combined to produce two probabilistic scores: one that identifies abnormal cell candidates (moderate, marked dysplasia, and cancer) and another that identifies normal bronchial epithelial cells to determine specimen adequacy. 3D images of the abnormal cell candidates were transmitted to a workstation for cytopathologist confirmation and final diagnosis, using a custom computer interface (Surveyor™). Figure 1

      Results
      Based on a stringent criterion for determining specimen adequacy, 12 of the 27 sputum specimens (or roughly 50%) were deemed adequate. In the adequate specimens, abnormal cells were automatically detected, correctly classified, and confirmed in 11 of the 12 cases. This result is statistically significant and demonstrates that the Cell-CT can achieve an abnormal cell detection rate for lung cancer cells in sputum nearing 92% sensitivity. The area under the receiver operator characteristic curve (aROC) for abnormal cell detection is 0.99 (see figure). We can estimate the specificity for normal specimens by examining the rate of correct normal cell detection and classification. With the detection of 9,683 normal cells, seven were falsely classified as abnormal (false positives). However, following human evaluation using Surveyor, the resulting specimen specificity was essentially 100%.

      Conclusion
      This first important study of Cell-CT shows there are sufficient abnormal cells in adequate, spontaneous sputa to achieve high-sensitivity lung cancer detection rates without false positives. The power of 3D single-cell morphometry supports the future potential impact of a non-invasive, sputum-based lung cancer screening test.