Researchers at The University of Hong Kong have developed MorphoGenie, a new artificial intelligence framework designed to improve disease diagnosis by interpreting complex cellular images. Led by Professor Kevin Tsia of the Department of Electrical and Computer Engineering, the team published their findings in Nature Communications in 2025. The study addresses a critical challenge in medical imaging: while cells contain vast amounts of health information, extracting reliable data from microscope images is difficult because subtle disease indicators are often invisible to the naked eye and existing AI tools function as opaque black boxes. MorphoGenie distinguishes itself by being interpretable, allowing researchers to understand not only the AI's predictions but also the specific visual features used to reach those conclusions. Inspired by human learning, the system utilizes the principle of compositionality to learn a small set of reusable visual building blocks from cell images. These building blocks include cell size and shape, broad internal textures, and fine local details. By recombining these elements, the tool can describe a wide range of cellular states and conditions without relying heavily on manual annotation or predefined assumptions. In testing, the HKU team demonstrated that MorphoGenie could successfully distinguish major lung cancer subtypes, detect drug-induced changes in cell morphology, and track dynamic biological processes such as cell-cycle progression and epithelial-to-mesenchymal transition. Dr. Rashmi Sreeramachandra Murthy, the study's first author, noted that cell images hold richer information than conventional measurements can capture, and MorphoGenie reveals biologically meaningful patterns that might otherwise remain hidden. A key strength of the framework is its versatility across different microscopy techniques, including label-free quantitative phase imaging and fluorescence microscopy. Furthermore, the system can transfer knowledge learned from one dataset to new, unseen data, suggesting broad applicability in drug discovery and studies of cellular responses to treatments. This adaptability supports the next generation of biomedical AI tools that prioritize transparency alongside performance. As artificial intelligence increasingly handles complex scientific tasks, the demand for systems that provide verifiable insights is growing. Professor Tsia emphasized that interpretability is essential for trust and scientific utility. If AI is to effectively aid in identifying meaningful cellular changes, its findings must be presented in a way that researchers can understand and verify. While fully autonomous biomedical discovery remains a future goal, MorphoGenie establishes a foundational approach for creating AI systems that are both powerful and transparent. By keeping human expertise at the center of the discovery process, this technology aims to advance the understanding of health and disease through more reliable and explainable diagnostics.