A research group led by Prof. WEI Yanjie from the Shenzhen Institute of Advanced Technology (SIAT) of the Chinese Academy of Sciences and their collaborators have proposed an effective deep learning architecture that enables detection and tracking, utilizing incomplete initial annotations. 

The study was published in International Journal of Molecular Sciences on November 7.  

Microscopical cell analysis has emerged as a crucial research area in biology, necessitating the development of algorithms for efficient cell detection and tracking. However, the reliance on pixel-wise annotations in fully supervised tasks requires significant manual effort which is time-consuming and often requires professional personnel. 

To addresses this challenge, the researchers developed a novel weakly supervised method for automated cell detection, achieved by updating incomplete initial annotations through cell tracking on brand new Induced Pluripotent Stem (iPS) cell dataset, which comprises brightfield and fluorescence images during the early reprogramming stage.    

The proposed method was evaluated using two fields of the iPS cell dataset, along with the cell detection accuracy (DET) evaluation metric from the Cell Tracking Challenge (CTC) initiative, and it achieved 0.862 and 0.924 DET, respectively. Subsequently, the transferability was tested using the public dataset FluoN2DH-GOWT1, which was taken from CTC，contains two datasets with reference annotations. Researchers randomly removed parts of the annotations in each labeled data to simulate the initial annotations on the public dataset.   

Experimental results showed that, with labels that comprise 10% cell markers, the DET improved from 0.130 to 0.903 and 0.116 to 0.877. When trained with labels that comprise 60% cell markers, the performance was better than the model trained using the supervised learning method. 

As highlighted by Prof. WEI, the implications of this study resonate across both the scientific and practical realms of cell analysis. Scientifically, our proposed framework challenges the conventional dependence on extensive manual annotations, paving the way for more efficient and scalable methods for studying cell dynamics. The study's demonstration of competitive performance across diverse datasets underscores the potential of utilizing incomplete initial labels to glean insights into cellular behaviors, ultimately aiding researchers in comprehending complex biological processes.    

"From a practical viewpoint, the framework's development is deemed promising as it significantly reduces labor-intensive annotation efforts in cell detection and tracking tasks. This efficiency gain holds the potential to expedite research timelines and democratize large-scale analyses, enabling a broader spectrum of researchers to advance various fields of biomedicine," said Prof. WEI. 

The proposed methods are promising and open up new avenues for efficient cell analysis in various biological applications.