[P006] Poster

1:00 PM - 1:50 PM JST(4:00 AM - 4:50 AM UTC)

[P03-438] Microalgae Biotechnology: Deciphering Artificial Intelligence-Driven and Machine Learning Modelling

○Kuan Shiong Khoo¹ (1. Yuan Ze University (Taiwan))

Keywords:

Microalgae,Artificial Intelligence,Classification,Image Processing,Machine Learning

Growing demand for microalgae-derived biomolecules and sustainable bioprocessing has driven the need for rapid, non-destructive, and scalable analytical methods. This research presents an integrated artificial intelligence (AI) framework that combines digital image processing, machine learning (ML), and deep learning (DL) to enable real-time microalgae classification of Chlorella vulgaris FSP-E, Spirulina platensis, and Chlamydomonas reinhardtii along with the prediction of C-phycocyanin (CPC) concentration in Spirulina platensis. Across multiple studies, image-based models demonstrated a strong potential alternative to conventional extraction-based workflows. Convolutional neural networks (CNNs) achieved superior performance over classical ML models when relying solely on image information. While Support Vector Machine (SVM) and Artificial Neural Network (ANN) benefited from auxiliary parameters such as absorbance and cultivation time. To address overfitting and robustness, hybrid stacking-ensemble architectures integrating CNN feature extraction with SVM and XGBoost regressors were developed, achieving highly stable and accurate CPC predictions (R² ≈ 0.998) with reduced variance and improved generalisation. Datasets comprising over 11,000 images captured under diverse lighting conditions and acquisition devices demonstrated that smartphone-based imaging can achieve competitive performance relative to digital cameras, supporting low-cost and scalable deployment. Importantly, biomass images alone were sufficient to deliver reliable CPC predictions, eliminating the need for destructive extraction steps. In parallel, comprehensive image preprocessing, segmentation, along shape and texture feature engineering enabled high-accuracy microalgae species classification, with ML, DL, and cloud-based vision models achieving accuracies exceeding 97%. Collectively, these findings demonstrate that AI-driven image analysis can transform microalgae monitoring into a real-time, non-invasive, and environmentally sustainable digital system, supporting next-generation biorefineries, smart cultivation, and automated microalgae management.