JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

Deep learning-enabled detection of rare circulating tumor cell clusters in whole blood using label-free, flow cytometry

  • 0Department of Biomedical Engineering, Tufts University, Medford, MA, 02155, USA. irene.georgakoudi@tufts.edu.
Lab on a Chip +

|

March 8, 2024

|

March 8, 2024

View abstract on PubMed

Summary

This summary is machine-generated.

This study introduces a deep-learning approach for detecting circulating tumor cell clusters (CTCCs) in blood. This advanced method enables real-time, in vivo monitoring, offering a promising tool for early cancer metastasis detection and prognosis.

Area Of Science

  • Biomedical Engineering
  • Cancer Research
  • Machine Learning Applications

Background

  • Metastatic tumors significantly worsen patient survival rates.
  • Circulating tumor cell clusters (CTCCs) are critical indicators of metastatic progression.
  • Existing detection methods for CTCCs lack in vivo and real-time capabilities.

Purpose Of The Study

  • To develop and validate a deep-learning (DL) based Confocal Backscatter and Fluorescence Flow Cytometry (BSFC) model for detecting CTCCs in whole blood.
  • To assess the performance of DL-based BSFC for in vivo and real-time CTCC monitoring.
  • To explore the potential of DL-based BSFC for clinical applications and ex vivo isolation of CTCCs.

Main Methods

  • Implementation of a deep-learning model for peak detection and classification of CTCCs using BSFC data.
  • Evaluation of DL-based BSFC performance using metrics such as false alarm rate and Pearson correlation coefficient.
  • Validation of the model on artificial spiking studies and blood samples from different species and cancer types using transfer learning.

Main Results

  • The DL-based BSFC model achieved a low false alarm rate (0.78 events/min) and high correlation (0.943) with expected events.
  • Detection purity of 72% and sensitivity of 35.3% for homotypic and heterotypic CTCCs (≥2 cells) were achieved.
  • The model demonstrated sensitivity to varying CTCC numbers and maintained performance across different species and cancer types.

Conclusions

  • Deep-learning enhanced BSFC offers a robust, label-free method for detecting CTCCs in whole blood with potential for in vivo monitoring.
  • The established performance metrics and cross-species validation support the clinical translation of DL-based BSFC.
  • Further advancements in throughput could enable critical applications in clinical CTCC detection and ex vivo isolation.