<p>Microstructural characterization of materials is fundamental for predicting macroscopic properties; however, it is typically influenced by characterization techniques and user subjectivity. In terms of irradiated materials, electron microscopy reveals significant variations in helium-bubble characteristics owing to diverse imaging conditions and task-specific requirements, thus necessitating robust statistical analysis. This study presents a novel approach that combines deep-learning and machine-learning methodologies to address these challenges. We introduce an interactive image-segmentation technique that utilizes minimal user annotations to guide a model in identifying and segmenting regions of interest. To overcome the limitations of existing deep-learning methods in detecting helium nanobubbles, we implement advanced computer vision-based machine-learning algorithms. This user-guided interactive machine-learning framework enables the extraction of relevant helium bubbles customized to specific user requirements, thereby mitigating the potential biases inherent in previous model annotations. Our methodology demonstrates enhanced accuracy and adaptability in helium-bubble analysis within irradiated materials, thereby contributing to more precise microstructural characterizations. The proposed approach is applicable to a wide range of material-science applications, thus offering a more objective and user-centric method for analyzing complex microstructural features.</p>