Automatic detection of circulating tumor cells in darkfield microscopic images of unstained blood using boosting techniques.

Anca Ciurte,Rares Buiga,Cristina Selicean,Olga Soritau,Olorunseun Ogunwobi

doi:10.1371/journal.pone.0208385

Abstract

Circulating tumor cells (CTCs) are nowadays one of the most promising tumor biomarkers. It is well correlated with overall survival and progression-free survival in breast cancer, as well as in many other types of human cancer. In addition, enumeration and analysis of CTCs could be important for monitoring the response to different therapeutic agents, thus guiding the treatment of cancer patients and offering the promise of a more personalized approach. In this article, we present a new method that could be used for the automatic detection of CTC in blood, based on the microscopic appearance of unstained cells. The proposed method is based on the evaluation of image characteristics and boosting techniques. A dataset of 263 dark field microscopy images was constructed and used for our tests, containing blood spiked with three different types of tumor cells. An overall sensitivity of 92.87% and a specificity of 99.98% were obtained for the detection of CTC, performances which proved to be comparable to those obtained by human experts.

Highlights

Circulating tumor cells are migratory cells in the blood that allow malignant tumors to establish metastatic colonies distributed throughout the body
In [18], we proposed a method of classifying Circulating tumor cells (CTCs) using several image characteristics, such as histogram statistics, grayscale co-occurrence matrix, gray tone difference matrix, color staistics and some new radial features, a method validated by a high classification rate
In this paper we proposed a new label-free method for identifying the presence of tumor cells in the blood, based on specific morphological characteristics, through an automated machine learning process

Summary

Introduction

Circulating tumor cells are migratory cells in the blood that allow malignant tumors to establish metastatic colonies distributed throughout the body. CTC identification techniques illustrates the concept of “liquid biopsy” which, unlike conventional biopsy, involves a non-invasive procedure for taking certain tumor products from the blood, making it possible to extract as much diagnostic and predictive information as possible. Because of their close correlation with overall survival and progression-free survival, CTCs promise to become one of the leading tumor markers in the near future. Their early presence in the blood can be used to detect solid tumors, provided that the problem of low sensitivity is solved

Methods

Results

Conclusion