SpheroQuaNNt: U-Net–based volume estimation in tumor spheroid and organoid models for assessing preclinical treatment response
- 1Center for Translational Research in Oncology (LIM24), Instituto do Cancer do Estado de Sao Paulo (ICESP), Hospital das Clinicas da Faculdade de Medicina da Universidade de Sao Paulo (HCFMUSP), São Paulo 01246-000, Brazil, São Paulo, Brazil
- 2Comprehensive Center for Precision Oncology, Universidade de São Paulo, São Paulo, 01246- 000 Brazil
Abstract
Three-dimensional (3D) cell culture models, including tumor spheroids and organoids, more accurately recapitulate the structural and physiological complexity of tissues than conventional monolayer cultures, making them valuable platforms for cancer research and drug screening. Volume is a biologically relevant parameter for monitoring growth dynamics and treatment response; however, its quantification often relies on manual image analysis, which is labor-intensive, subjective, and prone to inter-operator variability. Here, we introduce SpheroQuaNNt, a web-based application for automated spheroid segmentation and volume estimation using a U-Net Convolutional Neural Network (CNN). SpheroQuaNNt enables analysis of heterogeneous morphologies through image-based segmentation and provides a web-based interface for volume measurement. The model was trained and validated using 615 microscopy images representing diverse 3D culture systems, including monoculture and hybrid spheroids generated from human and canine tumor cell lines and human embryonic kidney cells, as well as tumor organoids derived from human melanoma lesions. After 30 training epochs, the model achieved 99.35% accuracy during training and 98.56% on the validation dataset, with an average Dice Score of 0.9478 for supervised validation. The method accurately captured volume variations across heterogeneous spheroids, although those presenting marked density heterogeneity between the core and periphery with a dark peripheral margin remained more challenging. Estimated volumes showed strong agreement with measurements performed independently by two experts, supporting its reliability. Overall, SpheroQuaNNt provides a standardized solution for volumetric characterization of spheroids and organoids, reducing operator-dependent variability and supporting high-throughput drug screening and quantitative analysis of 3D cell culture models.
Graphical abstract:
Supplementary Information: The online version contains supplementary material available at https://doi.org/10.1007/s10616-026-01074-9.
Related
- Preclinical Study on Novel Treatment Regimen for Elderly Patients with Drug-Sensitive Pulmonary Tuberculosis
- Automated spheroid and organoid image analysis from rule-based methods to deep learning
- Graphene oxide-induced multiple iPSC spheroid formation for enhanced production of angiogenic exosomes accelerating tissue regeneration.
- Ascorbic Acid Potentiates Cisplatin-Induced Antitumor Activity in HeLa Cervical Cancer Cells: Oxidative Stress, Mitochondrial Dysfunction, Apoptosis, and 3D Spheroid Responses
- Minocycline-Mediated Inhibition of Microglial Activation in Neuroinflammation-Related Neuropsychiatric Disorders: Preclinical Promise and Clinical Limitations
BibTeX
@article{dovaizem2026spheroquannt,
title = {{SpheroQuaNNt: U-Net–based volume estimation in tumor spheroid and organoid models for assessing preclinical treatment response}},
author = {Nathalia Leal Dovaizem and Jean Carlos dos Santos da Luz and Alexis Germán Murillo Carrasco and Luciana Nogueira de Sousa Andrade},
journal = {Cytotechnology},
year = {2026},
month = oct,
doi = {10.1007/s10616-026-01074-9},
url = {https://doi.org/10.1007/s10616-026-01074-9},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - SpheroQuaNNt: U-Net–based volume estimation in tumor spheroid and organoid models for assessing preclinical treatment response
AU - Nathalia Leal Dovaizem
AU - Jean Carlos dos Santos da Luz
AU - Alexis Germán Murillo Carrasco
AU - Luciana Nogueira de Sousa Andrade
PY - 2026
DA - 2026/10/25
JO - Cytotechnology
DO - 10.1007/s10616-026-01074-9
UR - https://doi.org/10.1007/s10616-026-01074-9
AB - Three-dimensional (3D) cell culture models, including tumor spheroids and organoids, more accurately recapitulate the structural and physiological complexity of tissues than conventional monolayer cultures, making them valuable platforms for cancer research and drug screening. Volume is a biologically relevant parameter for monitoring growth dynamics and treatment response; however, its quantification often relies on manual image analysis, which is labor-intensive, subjective, and prone to inter-operator variability. Here, we introduce SpheroQuaNNt, a web-based application for automated spheroid segmentation and volume estimation using a U-Net Convolutional Neural Network (CNN). SpheroQuaNNt enables analysis of heterogeneous morphologies through image-based segmentation and provides a web-based interface for volume measurement. The model was trained and validated using 615 microscopy images representing diverse 3D culture systems, including monoculture and hybrid spheroids generated from human and canine tumor cell lines and human embryonic kidney cells, as well as tumor organoids derived from human melanoma lesions. After 30 training epochs, the model achieved 99.35% accuracy during training and 98.56% on the validation dataset, with an average Dice Score of 0.9478 for supervised validation. The method accurately captured volume variations across heterogeneous spheroids, although those presenting marked density heterogeneity between the core and periphery with a dark peripheral margin remained more challenging. Estimated volumes showed strong agreement with measurements performed independently by two experts, supporting its reliability. Overall, SpheroQuaNNt provides a standardized solution for volumetric characterization of spheroids and organoids, reducing operator-dependent variability and supporting high-throughput drug screening and quantitative analysis of 3D cell culture models. Graphical abstract: Supplementary Information: The online version contains supplementary material available at https://doi.org/10.1007/s10616-026-01074-9.
SN - 0920-9069
SN - 1573-0778
ER -