Abstract:Assessing the quality of bovine satellite cells (BSCs) is essential for the cultivated meat industry, which aims to address global food sustainability challenges. This study aims to develop a label-free method for predicting fluorescence markers in isolated BSCs using deep learning. We employed a U-Net-based CNN model to predict multiple fluorescence signals from a single bright-field microscopy image of cell culture. Two key biomarkers, DAPI and Pax7, were used to determine the abundance and quality of BSCs. The image pre-processing pipeline included fluorescence denoising to improve prediction performance and consistency. A total of 48 biological replicates were used, with statistical performance metrics such as Pearson correlation coefficient and SSIM employed for model evaluation. The model exhibited better performance with DAPI predictions due to uniform staining. Pax7 predictions were more variable, reflecting biological heterogeneity. Enhanced visualization techniques, including color mapping and image overlay, improved the interpretability of the predictions by providing better contextual and perceptual information. The findings highlight the importance of data pre-processing and demonstrate the potential of deep learning to advance non-invasive, label-free assessment techniques in the cultivated meat industry, paving the way for reliable and actionable AI-driven evaluations.

What problem does this paper attempt to address?

### What problems does this paper attempt to solve? This paper aims to solve the challenges of evaluating the quality of bovine satellite cells (BSCs) in the cultured meat industry. Specifically, the goal of the research is to develop a label - free method to predict fluorescent markers from a single bright - field microscope image through deep - learning techniques, thereby achieving non - invasive quality assessment of BSCs. #### Main problems: 1. **Limitations of traditional methods**: - Currently, the evaluation of BSCs quality mainly depends on techniques such as immunofluorescence microscopy. These methods require the use of fluorescent dyes or antibodies, involve invasive sample preparation and expert annotation, increasing complexity and cost. - These methods provide limited contrast and specificity when dealing with complex biological samples and are difficult to cope with differences in cell isolation and culture conditions among different donors. 2. **Improving evaluation efficiency and accuracy**: - The research hopes to use deep - learning models (such as the convolutional neural network (CNN) based on the U - Net architecture) to predict multiple fluorescent signals, thereby reducing the need for invasive staining and accelerating the cell quality assessment process. - In particular, for two key biomarkers, DAPI (4',6 - diamidino - 2 - phenylindole) and Pax7 (paired box protein 7), which are used to determine the quantity and quality of BSCs. 3. **Coping with biological heterogeneity**: - There is significant heterogeneity in BSCs during proliferation and differentiation, which challenges the prediction ability of the model. The research improves data pre - processing (such as fluorescence denoising) and enhances visualization techniques (such as color mapping and image overlay) to improve the consistency and interpretability of model predictions. #### Solutions: - **Deep - learning model**: Adopt a U - Net - based CNN model and train it to predict fluorescent markers from bright - field microscope images. - **Data pre - processing**: Introduce a fluorescence denoising step to remove background noise and improve prediction performance and consistency. - **Enhanced visualization**: Through techniques such as color mapping and image overlay, improve the interpretability and perceived quality of prediction results. In short, this research aims to provide an efficient and reliable BSCs quality assessment tool for the cultured meat industry through a label - free, non - invasive deep - learning method.

Label-free prediction of fluorescence markers in bovine satellite cells using deep learning

Deep learning-based image analysis for in situ microscopic imaging of cell culture process

High-quality Label-Free Prediction of Fluorescence Images Through Stimulated Raman Scattering Imaging and Recurrent Deep Neural Network.

Stain-free approach to determine and monitor cell heath using supervised and unsupervised image-based deep learning

Evaluation of a Deep Learning Based Approach to Computational Label Free Cell Viability Quantification

Predicting fluorescent labels in label-free microscopy images with pix2pix and adaptive loss in Light My Cells challenge

Morphology-Based Deep Learning Approach for Predicting Osteogenic Differentiation

Label-free prediction of cell painting from brightfield images

Label-Free CD34+ Cell Identification Using Deep Learning and Lens-Free Shadow Imaging Technology

Fast-Training Deep Learning Algorithm for Multiplex Quantification of Mammalian Bioproduction Metabolites via Contactless Short-Wave Infrared Hyperspectral Sensing

Predictive Analysis of Linoleic Acid in Red Meat Employing Advanced Ensemble Models of Bayesian and CNN-Bi-LSTM Decision Layer Fusion Based Hyperspectral Imaging

Deep Learning in Label-free Cell Classification

Deep Learning-Based Automated Cell Detection-Facilitated Meat Quality Evaluation

Label-Free and in Situ Identification of Cells Via Combinational Machine Learning Models.

Prediction of Stem Cell State Using Cell Image‐Based Deep Learning

Cell culture product quality attribute prediction using convolutional neural networks and Raman spectroscopy

Learning deep features for dead and living breast cancer cell classification without staining

Robust virtual staining of landmark organelles

Label-Free Virtual HER2 Immunohistochemical Staining of Breast Tissue using Deep Learning

An Artificial Intelligence-Based Stacked Ensemble Approach for Prediction of Protein Subcellular Localization in Confocal Microscopy Images

Evaluation of deep learning training strategies for the classification of bone marrow cell images