AIM-CICs: an automatic identification method for cell-in-cell structures based on convolutional neural network

Area of Science:

• Cell Biology
• Computational Biology
• Biotechnology

Background:

• Current entosis research relies on manual quantification of cell-in-cell structures (CICs), which is time-consuming and oversimplified.
• Manual CIC quantification hinders in-depth mechanistic studies of entosis, a form of non-autonomous cell death.

Purpose of the Study:

• To develop an automated method for identifying and analyzing cell-in-cell structures (CICs) in entosis.
• To overcome the limitations of manual quantification and enable comprehensive CIC analysis.

Main Methods:

• Utilized artificial intelligence, specifically a convolutional neural network algorithm, to create an automated identification method for CICs (AIM-CICs).
• Trained the AIM-CICs model to differentiate between CICs and non-CICs and categorize CICs into five subclasses based on stage and cell number.

Main Results:

• The AIM-CICs demonstrated high accuracy in differentiating CICs from non-CICs (AUC > 0.99).
• AIM-CICs accurately categorized CICs into five subclasses with AUC > 0.97 for all categories.
• The developed method provides automated and efficient comprehensive CIC analysis.

Conclusions:

• The automated identification method for CICs (AIM-CICs) significantly advances entosis research by enabling efficient and detailed analysis.
• AIM-CICs facilitates high-throughput screening and profound mechanistic investigations into CIC-mediated cell death.