Fundamental Toxicological Sciences

---

An in vitro microfluidic culture device for peripheral neurotoxicity prediction at low concentrations based on deep learning

 

Xiaobo Han ¹, Naoki Matsuda ¹, Kazuki Matsuda ¹, Makoto Yamanaka ², Ikuro Suzuki <sup>1 

1</sup> Tohoku Institute of Technology (Japan) 
² Ushio Inc. (Japan) 

 

We developed a microfluidic culture device and an evaluation method using Deep Learning for a rapid assessment platform for peripheral neuropathy caused by anticancer drugs. Initially, primary rat dorsal root ganglia were cultured in the microfluidic culture device, separating the cell body and neurites. Immunofluorescence imaging was then used to analyze morphological changes in the neurites. Culturing of separated neurites in the microfluidic device for over a month demonstrated the viability of the evaluation process, including culture, drug administration, and observation with fluorescent microscopy. 

We also conducted model experiments for drug testing using anticancer drugs known to cause peripheral neurotoxicity (vincristine, oxaliplatin, and paclitaxel). The results showed that the trained AI could identify morphological changes in the neurites caused by each compound and accurately predict toxicity even at low concentrations. Additionally, the AI was able to accurately detect neurotoxicity based on neurite images, even for untrained compounds at low concentrations. These results suggest that this microfluidic culture system and evaluation method are useful for in vitro toxicity assessment. 

 

For papers (You are leaving for an external site.)


 

Copyright © USHIO INC. All Rights Reserved.