Redefining circulating tumor cells by image processing

Abstract

Circulating tumor cells (CTC) in the blood of patients with metastatic carcinomas are associated with poor survival and can be used to guide therapy. However, CTC are very heterogeneous in size and shape, and are present at very low frequencies. Missing or misjudging a few events may have great consequences, as the clinically used cut-off for determining favorable or unfavorable prognosis for a patient is currently at three or five CTC in 7.5 ml of blood, depending on the cancer type. Current qualitative methods rely on human classification of CTC and suffer from intra- and inter-reviewer variability. We show that it is possible to define CTC in breast, colon, and prostate cancer by computerized algorithms with the same prognostic value as human reviewing, by using patient survival data as the training parameter. Our new definition revealed that morphological parameters, such as size of CTC, differ between cancer types and are correlating with survival. Furthermore, we show that patients should be treated until the number of CTC is zero and that other definitions of CTC, allowing for example tumor cell fragments to be counted, do not improve the prognostic value of the method. Our computerized definition enabled quantitative measurements of biomarkers, such as Her-2, in CTC and showed great heterogeneity of such biomarkers within patients. In conclusion, we have created and validated automated CTC characterization, which eliminates reviewer variability, is fast and low cost, and allows for quantitative biomarker assessment in the future. Measuring expression of these receptors on CTC may allow for a true quantitative liquid biopsy. Such a biopsy, which is less tedious and hazardous to collect from patients than a traditional biopsy, could improve cancer diagnostics significantly.