2018/07/30 by Yaroslav Shmulev, Shmulev, Yaroslav, Mikhail Belyaev +1
Computer Science · Mathematics · Medicine · Neuroscience · Psychology · #Artificial intelligence #Brain Tumor Detection and Classification #Cognition #Computer Vision and Pattern Recognition (cs.CV) #Computer science #Deep learning #Dementia and Cognitive Impairment Research #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Machine Learning in Healthcare #Machine learning #Neuroimaging #Psychiatry #Psychology #cs.CV #cs.LG #stat.ML
paper · pdf · doi:10.48550/arxiv.1807.11228
published in arXiv (Cornell University) (Cornell University)
arxiv created 2018/07/30 · openalex publication_date 2018/07/30 · arxiv updated 2018/07/31 · openalex created_date 2022/08/04 · openalex updated_date 2026/08/04
Nowadays, a lot of scientific efforts are concentrated on the diagnosis of\nAlzheimer's Disease (AD) applying deep learning methods to neuroimaging data.\nEven for 2017, there were published more than a hundred papers dedicated to AD\ndiagnosis, whereas only a few works considered a problem of mild cognitive\nimpairments (MCI) conversion to the AD. However, the conversion prediction is\nan important problem since approximately 15% of patients with MCI converges to\nthe AD every year. In the current work, we are focusing on the conversion\nprediction using brain Magnetic Resonance Imaging and clinical data, such as\ndemographics, cognitive assessments, genetic, and biochemical markers. First of\nall, we applied state-of-the-art deep learning algorithms on the neuroimaging\ndata and compared these results with two machine learning algorithms that we\nfit using the clinical data. As a result, the models trained on the clinical\ndata outperform the deep learning algorithms applied to the MR images. To\nexplore the impact of neuroimaging further, we trained a deep feed-forward\nembedding using similarity learning with Histogram loss on all available MRIs\nand obtained 64-dimensional vector representation of neuroimaging data. The use\nof learned representation from the deep embedding allowed to increase the\nquality of prediction based on the neuroimaging. Finally, the current results\non this dataset show that the neuroimaging does affect conversion prediction,\nhowever, cannot noticeably increase the quality of the prediction. The best\nresults of predicting MCI-to-AD conversion are provided by XGBoost algorithm\ntrained on the clinical and embedding data. The resulting accuracy is 0.76 +-\n0.01 and the area under the ROC curve - 0.86 +- 0.01.\n