Application of Machine Learning for Label-Free Detection of Circulating Tumor Cell Clusters in Whole Blood Using Flow Cytometry.
Vora, Nilay.
Shekhar, Prashant.
Esmail, Michael.
Patra, Abani.
Georgakoudi, Irene.
2022
-
Circulating tumor cells (CTCs) and circulating tumor cell clusters (CTCCs) are rare cellular events found in the blood stream of cancer patients. CTCCs are associated with increased metastatic potential and a reduced overall survival rate. To date, the only FDA approved method for the detection of CTCs and CTCCs is CellSearch®. However, detection of CTCCs by CellSearch® is hampered by the infrequency ... read moreof cluster occurrence in the blood stream. With less than 4 CTCCs/7.5mL of blood, CellSearch® has poor correlation with prognosis, necessitating the development of new method for detection and isolation of CTCCs. Here we utilize a custom confocal back scatter and fluorescence flow cytometer (CBSFFC) to optically detect CTCCs in whole blood using endogenous scattering and machine learning methods in vitro. Whole blood samples were isolated from mice or rats and spiked with GFP expressing MDA-MB-231 cancer cells. GFP expression was used as a ground truth label to discern location of CTCCs from blood scatter events. Over the course of 18 independent days, we established a dataset composed of over 6000 CTCC peaks and 60000 Non-CTCC (NC) peaks. Thirteen days of data were used for training with the remaining 5 days being used to evaluate performance. To determine the optimal model for classifying CTCC peaks from NC peaks, we trained and evaluated a narrow neural network, k-nearest neighbor model, and ensemble boosted tree model. After optimization, we demonstrated that the use of an ensemble boosted tree model yields the greatest overall performance. The comprehensive classification algorithm demonstrated greater than 90% sensitivity, specificity, and accuracy, and a purity greater than 80%. These results indicate CTCCs feature unique scattering signatures which can be discerned from NC scattering, motivating further development of in vivo and in vitro methods of label-free CTCC detection using machine learning methods.
15-minute talk presented by Nilay Vora at the Tufts Graduate Student Council's 27th Graduate Student Research Symposium, April 20, 2022, 3rd place winner.read less - Vora, Nilay, Prashant Shekhar, Michael Esmail, Abani Patra, Irene Georgakoudi. "Application of Machine Learning for Label-Free Detection of Circulating Tumor Cell Clusters in Whole Blood Using Flow Cytometry." Presentation at the 27th Graduate Research Symposium, Tufts University, April 20, 2022.
- ID:
- wh247760s
- To Cite:
- TARC Citation Guide EndNote
- Usage:
- Detailed Rights
