The Global Kernel k-means Clustering Algorithm for Cerebral Infarction Classification

摘要

Cerebral infarction is the death of neurons, glia cells and blood vessel systems caused by a lack of oxygen and nutrients. This situation is often called storke. Common causes of neuron damage are hypoxia, which is caused by impaired blood flow, reduced oxygen pressure in blood circulation, toxins, and hypoglycemia which can result in the same morphological changes as morphological changes in hypoxia. Hypoxia is reduced oxygen pressure in the alveoli, resulting in hypoxemia which can cause hypoxic brain tissue. The initial stage of ischemic neurons is characterized by the formation of micro vacuolization, which is characterized by the size of the cells that are still normal or slightly reduced, the nucleus shrinks slightly, vacuoles occur in the perikaryon region. This micro vacuole can be found in neurons in hippokamus and cortical 5-15 minutes after hypoxia. The final sign of cell damage due to ischemia is characterized by the nucleus becoming pyknotic and fragmented. To classify cerebral infarction, the author uses the global k-means clustering algorithm as a classification method that shows that the method has good accuracy, good memory, and good precision in classifying cerebral infarction. In this proposed method, the global kernel k-means clustering algorithm is an extension of the standard k-means clustering algorithm and has been used to identify or classify clusters that are non-linearly separated in space input. This method adds one cluster at each stage through a global search process consisting of several k-means kernel executions from the appropriate initialization. Therefore, this method can make good classification accuracy. In particular, this achieves classification accuracy of up to 78% for the highest accuracy.