Neural Network Training with Highly Incomplete Datasets published in Machine Learning: Science and Technology

Working principles for training neural networks with highly incomplete dataset: vanilla (upper panel) vs GapNet (lower panel) (Image by Yu-Wei Chang.)
Neural Network Training with Highly Incomplete Datasets
Yu-Wei Chang, Laura Natali, Oveis Jamialahmadi, Stefano Romeo, Joana B. Pereira, Giovanni Volpe
Machine Learning: Science and Technology 3, 035001 (2022)
arXiV: 2107.00429
doi: 10.1088/2632-2153/ac7b69

Neural network training and validation rely on the availability of large high-quality datasets. However, in many cases only incomplete datasets are available, particularly in health care applications, where each patient typically undergoes different clinical procedures or can drop out of a study. Since the data to train the neural networks need to be complete, most studies discard the incomplete datapoints, which reduces the size of the training data, or impute the missing features, which can lead to artefacts. Alas, both approaches are inadequate when a large portion of the data is missing. Here, we introduce GapNet, an alternative deep-learning training approach that can use highly incomplete datasets. First, the dataset is split into subsets of samples containing all values for a certain cluster of features. Then, these subsets are used to train individual neural networks. Finally, this ensemble of neural networks is combined into a single neural network whose training is fine-tuned using all complete datapoints. Using two highly incomplete real-world medical datasets, we show that GapNet improves the identification of patients with underlying Alzheimer’s disease pathology and of patients at risk of hospitalization due to Covid-19. By distilling the information available in incomplete datasets without having to reduce their size or to impute missing values, GapNet will permit to extract valuable information from a wide range of datasets, benefiting diverse fields from medicine to engineering.

Multiplex Connectome Changes across the Alzheimer’s Disease Spectrum Using Gray Matter and Amyloid Data published in Cerebral Cortex

Brain nodes. (Image taken from the article.)
Multiplex Connectome Changes across the Alzheimer’s Disease Spectrum Using Gray Matter and Amyloid Data
Mite Mijalkov, Giovanni Volpe, Joana B Pereira
Anna Canal-Garcia, Emiliano Gómez-Ruiz, Mite Mijalkov, Yu-Wei Chang, Giovanni Volpe, Joana B Pereira, Alzheimer’s Disease Neuroimaging Initiative
Cerebral Cortex, bhab429 (2022)
doi: 10.1093/cercor/bhab429

The organization of the Alzheimer’s disease (AD) connectome has been studied using graph theory using single neuroimaging modalities such as positron emission tomography (PET) or structural magnetic resonance imaging (MRI). Although these modalities measure distinct pathological processes that occur in different stages in AD, there is evidence that they are not independent from each other. Therefore, to capture their interaction, in this study we integrated amyloid PET and gray matter MRI data into a multiplex connectome and assessed the changes across different AD stages. We included 135 cognitively normal (CN) individuals without amyloid-β pathology (Aβ−) in addition to 67 CN, 179 patients with mild cognitive impairment (MCI) and 132 patients with AD dementia who all had Aβ pathology (Aβ+) from the Alzheimer’s Disease Neuroimaging Initiative. We found widespread changes in the overlapping connectivity strength and the overlapping connections across Aβ-positive groups. Moreover, there was a reorganization of the multiplex communities in MCI Aβ + patients and changes in multiplex brain hubs in both MCI Aβ + and AD Aβ + groups. These findings offer a new insight into the interplay between amyloid-β pathology and brain atrophy over the course of AD that moves beyond traditional graph theory analyses based on single brain networks.

Yu-Wei Chang visits the Soft Matter Lab. Welcome!

Yu-Wei Chang is an engineer at the Digital Medicine Center at National Yang-Ming University in Taiwan, working on a machine learning approach for phenotyping psychiatric disorders.
He will visit us for 2 months from February 29, 2020, to April 30, 2020, and he will be working on deep learning for Alzheimer’s Disease as well as the development of BRAPH 2.0 (March-April 2020).