Self-supervised learning allows precise positioning of onion-like carbon nanoparticles in colorectal cancer cells

Onion-like carbon nanoparticles (OLC) have shown great potential in photothermal cancer therapy, but precise, real-time tracking of their distribution within living cells is crucial for optimizing their therapeutic effects. This study introduces a groundbreaking method combining limited-angle digital holographic tomography (DHT) with self-supervised learning to track the three-dimensional distribution of OLC nanoparticles in colorectal cancer cells (CRCs). We developed an internal learning neural network (ILNN) to enhance phase image reconstruction at unmeasured angles, addressing the data limitations of conventional methods. After validating this technique with SiO2 microspheres, we applied it to monitor OLC nanoparticle distribution in CRC cells over a 2-hour period. By quantifying changes in the surface area and volume of nanoparticles, we gained valuable insights into their temporal evolution. This innovative approach enables non-invasive, dynamic monitoring of nanoparticles in living cells, paving the way for improved photothermal therapy strategies and more effective treatment models in cancer management.