Using CNN with Bayesian optimization to identify cerebral micro-bleeds

This article studies the problem of detecting cerebral micro-bleeds (CMBs) using a convolutional neural network (CNN). Cerebral micro-bleeds (CMBs) are increasingly recognized neuroimaging findings, occurring with cerebrovascular diseases, dementia, and normal aging. Naturally enough, it becomes necessary to detect CMBs in the early stages of life. The focus of this article is to infuse new techniques like Bayesian optimization to find the optimum set of hyper-parameters efficiently, making even the simplest of CNN architectures perform well on the problem. Experimentally, we observe our CNN (five layers, i.e., two convolution, two pooling, and one fully connected) achieves accuracy = 98.97%, sensitivity = 99.66%, specificity = 98.14%, and precision = 98.54% on the test set (hold-out validation) when calculated over an average of ten runs. The proposed model outperformed state-of-the-art methods.