Fig. 5: Extracted feature via t-SNE visualizations of feature spaces from different architectures. | npj Biosensing

Fig. 5: Extracted feature via t-SNE visualizations of feature spaces from different architectures.

From: Accurate label free classification of cancerous extracellular vesicles using nanoaperture optical tweezers and deep learning

Fig. 5

a TrapNet, b LSTM, c Bi-LSTM, d CNN, e CLSTM, and f Transformer, applied to classify three EV types: MDA-MB-231 (purple), MCF7 (green), and MCF10A (yellow). These plots show the clustering of individual EVs based on extracted features, with color indicating EV class. The degree of cluster separation reflects each model’s ability to extract features and classify EVs accurately.

Back to article page