Deep learning identifies cell death types in ischemia reperfusion injury

After blood flow is restored to the brain, stroke patients can still suffer additional damage as oxygen surges into injured tissue—a condition called cerebral ischemia-reperfusion injury (CIRI). Researchers at Huaihe Hospital of Henan University say a deep learning system can automatically classify how brain cells die in CIRI, distinguishing autophagy, apoptosis and necrosis, tasks that typically rely on slow, subjective microscope review. The work, published as open access in BMC Biology, reports accuracy exceeding 93%. The study—led by Peng Zhang, Chenyang Xu, Zhijun Liu and Yadong Geng, with Honglin Liu as corresponding author—built a ResNet-based convolutional neural network trained on standardized microscope images from animal models.