Human-in-the-loop deep learning for scalable detection of circulating tumor cells and cancer-associated fibroblasts by imaging flow cytometry: a dual-center validation study

Abstract Circulating tumor cells and cancer-associated fibroblasts (CTCs/cCAFs) are promising biomarkers, but their rarity and heterogeneity make identification labor-intensive, even with standardized technologies such as imaging flow cytometry (imFC). Here, we developed a human-in-the-loop deep-learning workflow addressing this bottleneck by prioritizing candidate cells for expert review. A model trained on five-channel imFC-images and expert-annotated breast cancer reference CTCs/cCAFs libraries was evaluated in 59 patients with 5 cancer types and 6 healthy donors at two centers/devices. In peripheral blood, the model detected 443/490 CTCs/cCAFs among 7.8 million events, with 54.8 false positives/100,000 events. Reviewing 50 prioritized candidates/sample reached the predefined threshold in eligible samples, taking 2min 15sec/sample versus an estimated 30min manually. In tumor-draining vein blood, 1,200/1,678 CTCs/cCAFs were detected. Detection of ≥1CTC was associated with lymph node and/or distant metastases (p=0.016). This workflow may facilitate scalable imFC analysis with expert verification. Multicenter studies are needed to confirm generalizability and clinical relevance.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-17
DOI
https://doi.org/10.21203/rs.3.rs-11677722/v1
Primary Topic
Cancer Genomics and Diagnostics
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Human-in-the-loop deep learning for scalable detection of circulating tumor cells and cancer-associated fibroblasts by imaging flow cytometry: a dual-center validation study

Marcin Matuszewski, Adam Durczyński, Kevin Miszewski, Natalia Bednarz‐Knoll et al.
1 links
Research Square
Cancer Genomics and Diagnostics
preprint

Human-in-the-loop deep learning for scalable detection of circulating tumor cells and cancer-associated fibroblasts by imaging flow cytometry: a dual-center validation study

Marcin Matuszewski, Adam Durczyński, Kevin Miszewski, Natalia Bednarz‐Knoll, Jarek Kobiela, Rafał Dziadziuszko, Krystian Zielinski, Jędrzej Świeżewski, Anna Joanna Zaczek, Anna Muchlińska, Aleksandra Parteka, Elżbieta Senkus; Katedra i Klinika Onkologii i Radioterapii, Gdański Uniwersytet Medyczny; Uniwersyteckie Centrum Kliniczne w Gdańsku Centrum Chorób Piersi, Agnieszka Piastowska-Ciesielska, Wojciech Ciesielski
preprint en

Abstract

Abstract Circulating tumor cells and cancer-associated fibroblasts (CTCs/cCAFs) are promising biomarkers, but their rarity and heterogeneity make identification labor-intensive, even with standardized technologies such as imaging flow cytometry (imFC). Here, we developed a human-in-the-loop deep-learning workflow addressing this bottleneck by prioritizing candidate cells for expert review. A model trained on five-channel imFC-images and expert-annotated breast cancer reference CTCs/cCAFs libraries was evaluated in 59 patients with 5 cancer types and 6 healthy donors at two centers/devices. In peripheral blood, the model detected 443/490 CTCs/cCAFs among 7.8 million events, with 54.8 false positives/100,000 events. Reviewing 50 prioritized candidates/sample reached the predefined threshold in eligible samples, taking 2min 15sec/sample versus an estimated 30min manually. In tumor-draining vein blood, 1,200/1,678 CTCs/cCAFs were detected. Detection of ≥1CTC was associated with lymph node and/or distant metastases (p=0.016). This workflow may facilitate scalable imFC analysis with expert verification. Multicenter studies are needed to confirm generalizability and clinical relevance.

Research Square
Medical University of Lodz (PL), Gdańsk Medical University (PL)
Good health and well-being
Cancer Genomics and Diagnostics
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon