Revisiting Cancer Cell Polarity Through Artificial Intelligence: A Perspective on Image-Based Biomarkers for Precision Oncology

Cancer research has traditionally emphasized genetic and molecular alterations, while fundamental biophysical characteristics such as cell polarity have received comparatively limited attention in diagnostic and therapeutic contexts. Cell polarity plays a critical role in normal tissue architecture, function, and homeostasis, and its disruption is a well-recognized feature of malignant transformation, invasion, and metastasis [1–4]. Despite its biological significance across both healthy and diseased tissues, polarity is rarely quantified or systematically incorporated into clinical workflows. This perspective proposes that recent advances in biomedical imaging and artificial intelligence (AI) provide a timely opportunity to reposition cell polarity as a meaningful, image-derived biomarker in cancer.

We outline a conceptual framework in which AI-enabled image analysis is applied to high-resolution microscopy and digital pathology images to assess polarity-related spatial organization within cells. Machine learning and deep learning approaches, particularly convolutional neural networks, have demonstrated strong capability in extracting complex spatial features from biomedical images [5,6]. Within this framework, AI models could be leveraged to quantify asymmetries in cytoskeletal organization, nuclear positioning, and intracellular feature distribution, enabling objective and scalable polarity assessment beyond conventional morphology-based evaluation.

From a conceptual standpoint, AI-driven polarity analysis has the potential to reveal phenotypic information associated with tissue organization, tumour aggressiveness, metastatic behaviour and therapeutic responsiveness. Prior work in computational pathology demonstrates that deep learning can uncover clinically relevant patterns not readily discernible to human observers [7]. Polarity-associated spatial signatures may therefore complement molecular and histopathological data, contributing to more nuanced cancer classification and risk stratification strategies.

In conclusion, cell polarity represents an underexplored yet biologically grounded target for AI-driven image analysis. Integrating spatial biology with AI-based imaging analytics may enable more holistic, interpretable, and data-driven approaches to cancer diagnosis, prognosis, and treatment evaluation in precision oncology.