Object Tracking in a $360^o$ View: A Novel Perspective on Bridging the Gap to Biomedical Advancements

TL;DR

The paper surveys the landscape of object tracking with a strong emphasis on biomedical imaging, outlining traditional, probabilistic, feature-based, and deep learning paradigms. It synthesizes how detection, classification, and temporal modeling—via tools like CNNs, LSTMs, Transformers, and end-to-end frameworks such as MOTR—address challenges from occlusion to multi-object interactions. Key contributions include a taxonomy of methods, a critical appraisal of their six desiderata (Extensiveness, Robustness, Trainability, Multi-Domain Compatibility, End-to-End Functionality, Scalability), and a roadmap for next-generation, cross-domain tracking systems tailored to bio-imaging and cellular dynamics. The work highlights trends such as self-supervised representation learning, contrastive coding, attention-based models (e.g., SAMURAI), GAN-enhanced tracking, and deep RL, demonstrating their potential to transform spatiotemporal analysis in biology and medicine. Overall, the paper argues that integrated, end-to-end, cross-domain tracking systems—capable of handling 2D/3D data, diverse imaging modalities, and real-time demands—will accelerate biomedical discovery and clinical translation.

Abstract

Object tracking is a fundamental tool in modern innovation, with applications in defense systems, autonomous vehicles, and biomedical research. It enables precise identification, monitoring, and spatiotemporal analysis of objects across sequential frames, providing insights into dynamic behaviors. In cell biology, object tracking is vital for uncovering cellular mechanisms, such as migration, interactions, and responses to drugs or pathogens. These insights drive breakthroughs in understanding disease progression and therapeutic interventions. Over time, object tracking methods have evolved from traditional feature-based approaches to advanced machine learning and deep learning frameworks. While classical methods are reliable in controlled settings, they struggle in complex environments with occlusions, variable lighting, and high object density. Deep learning models address these challenges by delivering greater accuracy, adaptability, and robustness. This review categorizes object tracking techniques into traditional, statistical, feature-based, and machine learning paradigms, with a focus on biomedical applications. These methods are essential for tracking cells and subcellular structures, advancing our understanding of health and disease. Key performance metrics, including accuracy, efficiency, and adaptability, are discussed. The paper explores limitations of current methods and highlights emerging trends to guide the development of next-generation tracking systems for biomedical research and broader scientific domains.

Figures (35)

• Figure 1: Sample cell segmentation and tracking in mitochondria of lung cell andT. gondii. The top row indicates the mitochondria in an original frame and the segmented one across time in the top right. The cells are segmented using the watershed algorithm and connected component labeling to see how many connected components we have over time and how they are changing along with the video. The bottom row illustrates sample tracking ofT. gondiiusing segmentation and object association.
• Figure 2: Cheng’s Taxonomy of methods for object detection in optical RSIs. The components are discussed in his paper b46.
• Figure 3: Contour-based tracking applied onT. gondiicell-tracking by setting a bounding box around the detected contours of the cells.
• Figure 4: Contour-based tracking results in different video sequences. (A) Tennis player and (B) a girl playing on the balcony.
• Figure 5: KLT feature tracking applied on biological cells, highlighting the movement of distinct points in a time-lapse video. The green lines represent the trajectories of the tracked features.