Information Sciences in Imaging 2013 Seminar Series: April


Computerized Tomography (CT) of Single Cells for Early Cancer Detection: Application to Lung Cancer Diagnosis from Sputum


Nearly all cancer diagnoses are based on conventional microscopy with manual screening from 2D focal plane images. However, the cell is inherently a 3D object that cannot be accurately interrogated using 2D image analysis alone. Employing instruments such as confocal microscopy to generate a stack of 2D focal plane images, as an approximation of the 3D object, inevitably creates a non-spherical point spread function that results in feature measurements that are orientation dependent. 

A true 3D tomographic reconstruction of the cell overcomes the orientation dependence, and the feature measurements then become highly robust. We developed a new instrument, called the Cell-CT, to compute single-cell true-3D tomographic images. The Cell-CT is fully automated and does not require human interpretation. Analysis of these 3D images yields predictive and diagnostic features that represent a comprehensive assessment of cell phenotypic biosignatures that drive classifiers to achieve both high sensitivity and high specificity. 

The case of lung cancer diagnosis from cells in sputum, as a non-invasive screening test, will be presented. In human clinical studies we have demonstrated an ROC area of 0.98 with an operating point sensitivity of 97% and specificity of 99% for the early detection of lung adenocarcinoma. The potential impact of this breakthrough technology will be discussed in the context of patient outcomes and cost-effectiveness.