LuTuFlow: a deep learning-based pipeline for quantitative characterization of mouse lung tumors in CT imaging
- 1EPFL (Swiss Fed. Inst. of Technology Lausanne)
Abstract
Computed tomography (CT) is a widely used, non-invasive imaging tool to monitor tumor growth. Tumor identification and volumetric quantification are crucial for assessing tumor burden, whereas tracking individual tumors by aligning scans from multiple time points reveals disease progression. However, these processes remain largely dependent on manual analysis, limiting the accuracy, reproducibility, objectivity, and scalability of tumor characterization. Several deep learning models have enabled automated segmentation and object identification. However, open-source, integrated software platforms that cover the entire workflow - including processing raw CT images, annotating and tracking tumors through time, and generating data reports and visualizations - are still needed. Here, we present LuTuFlow, a computer vision pipeline incorporating a nnU-Net-based deep learning model, configured for murine lung segmentation and tumor analysis. We integrated these modules into a Napari plugin, allowing researchers to apply the entire automated workflow to transform raw CT scan datasets into quantitative readouts and interactive visualizations in a simple graphical user interface. The resulting model, iteratively refined using expert-corrected annotations, generated reliable quantitative estimates of tumor burden.
Related
- Quantitative super-resolution imaging with deep learning segmentation reveals early mitochondrial network remodeling and altered mitochondria-lysosome interactions in Pompe disease muscle
- Quantitative MRI Characterization of Fluid-Related Phenotypes in Pediatric Epilepsy: Deep Learning-Assisted Assessment of Enlarged Perivascular Space Burden, Regional Distribution, and Clinical Associations
- A three-dimensional imaging pipeline for the detection of vasculogenic mimicry in mouse tumor models
- Quantitative dual-energy computed tomography parameters as imaging biomarkers for assessing the histopathologic grade of colorectal cancer
- Relative CT Number as a Quantitative Imaging Biomarker for Differentiating Sinonasal Polyp, Inverted Papilloma, and Squamous Cell Carcinoma: A Preliminary Retrospective Study
BibTeX
@misc{liu2026lutuflow,
title = {{LuTuFlow: a deep learning-based pipeline for quantitative characterization of mouse lung tumors in CT imaging}},
author = {Y. Liu and L. Li and M. Wittwer and A. Martinez-Usatorre and N. Mansouri and B. Dapia and E. Ando and M. De Palma},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.12.758509},
url = {https://www.biorxiv.org/content/10.64898/2026.10.12.758509v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - LuTuFlow: a deep learning-based pipeline for quantitative characterization of mouse lung tumors in CT imaging
AU - Y. Liu
AU - L. Li
AU - M. Wittwer
AU - A. Martinez-Usatorre
AU - N. Mansouri
AU - B. Dapia
AU - E. Ando
AU - M. De Palma
PY - 2026
DA - 2026/10/12
PB - bioRxiv
DO - 10.64898/2026.10.12.758509
UR - https://www.biorxiv.org/content/10.64898/2026.10.12.758509v1
AB - Computed tomography (CT) is a widely used, non-invasive imaging tool to monitor tumor growth. Tumor identification and volumetric quantification are crucial for assessing tumor burden, whereas tracking individual tumors by aligning scans from multiple time points reveals disease progression. However, these processes remain largely dependent on manual analysis, limiting the accuracy, reproducibility, objectivity, and scalability of tumor characterization. Several deep learning models have enabled automated segmentation and object identification. However, open-source, integrated software platforms that cover the entire workflow - including processing raw CT images, annotating and tracking tumors through time, and generating data reports and visualizations - are still needed. Here, we present LuTuFlow, a computer vision pipeline incorporating a nnU-Net-based deep learning model, configured for murine lung segmentation and tumor analysis. We integrated these modules into a Napari plugin, allowing researchers to apply the entire automated workflow to transform raw CT scan datasets into quantitative readouts and interactive visualizations in a simple graphical user interface. The resulting model, iteratively refined using expert-corrected annotations, generated reliable quantitative estimates of tumor burden.
ER -