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

Abstract

Whereas biochemical markers are available for most types of cell death, current studies on non-autonomous cell death by entosis relays strictly on the identification of cell-in-cell structure (CICs), a unique morphological readout that can only be quantified manually at present. Moreover, the manual CICs quantification is generally over-simplified as CICs counts, which represents a major hurdle against profound mechanistic investigations. In this study, we take advantage of artificial intelligence (AI) technology to develop an automatic identification method for CICs (AIM-CICs), which performs comprehensive CICs analysis in an automated and efficient way. The AIM-CICs, developed on the algorithm of convolutional neural network (CNN), can not only differentiate between CICs and non-CICs (AUC > 0.99), but also accurately categorize CICs into five subclasses based on CICs stages and cell number involved (AUC > 0.97 for all subclasses). The application of AIM-CICs would systemically fuel researches on CICs-mediated cell death such as high-throughput screening.

Subtype-Based Analysis of Cell-in-Cell Structures in Esophageal Squamous Cell Carcinoma

“…Third, some samples came from the commercial TMA and the related treatment information including chemotherapy and radiotherapy were unavailable which may decrease the credibility of result to certain extents. Additionally, the quantification of CIC structures in tissues relays on multiple experienced investigators, which calls for an algorithm-based program for more standard and efficient quantification similar to that achieved recently on cytospins (48). TiT low, <3.721 CICs/mm 2 ; high, ≥3.721 CICs/mm 2 .…”

Molecular mechanisms underlying cell-in-cell formation: core machineries and beyond

“…Recent progress in image recognition by machine learning or artificial intelligence (AI) may shed light on this technical bottle neck. We recently made attempts to identify CICs by the algorithms of convolutional neural network and obtained ideal results in recognizing CICs on cytospin slides (specificity and sensitivity: >97%, respectively) ( Tang et al, 2021 ). Thus, it is expected that a high-content screening, combined with AI recognition of CICs, for biochemical or genetical regulators of CIC formation, would be a feasible way to speed up the mechanistic studies of CIC-mediated death.…”