Abstract
With the prevalence of global aging, the number of patients with neurodegenerative diseases and the cost of care are rising rapidly. It is necessary to change the way individuals are diagnosed and treated, and develop effective therapeutic interventions. In the past decade, due to the generation of new technologies, the data of neurodegenerative patients has accumulated rapidly, such as medical image data, omics data of biological samples, electronic medical record data, activation and behavior data, and social media data. Traditional data analysis methods can only find the change of a single variable or perform simple correlation, which have been unable to meet the needs of high-dimensional big health datasets. Machine learning, which can directly learn the rules from the input data and help provide biological insights, has been widely used to help diagnosis and prognosis for neurodegenerative diseases.
Although machine learning has shown great potential, a number of significant data-related problems plague the application of machine learning in helping diagnosis and prognosis of neurodegenerative diseases, such as excessive data noise, data (labeled data) deficiency, data heterogeneity, data imbalance, multi-source data, and data isolation. Most existing machine learning-based methods can only deal with high-quality, sufficient, homogeneous, class-balanced, single-source, and accessible data. In order to improve the applicability of machine learning in helping the diagnosis and prognosis of neurodegenerative diseases, it is necessary to develop new data preprocessing methods and data representation learning methods.
This research topic aims to provide a diverse but complementary set of contributions to demonstrate new developments and applications that cover existing issues in data processing and analysis of neurodegenerative diseases data for helping diagnosis and prognosis. We would also like to accept successful applications of the new methods, including but not limited to data processing, analysis, and knowledge discovery of neurodegenerative diseases data for helping diagnosis and prognosis
Although machine learning has shown great potential, a number of significant data-related problems plague the application of machine learning in helping diagnosis and prognosis of neurodegenerative diseases, such as excessive data noise, data (labeled data) deficiency, data heterogeneity, data imbalance, multi-source data, and data isolation. Most existing machine learning-based methods can only deal with high-quality, sufficient, homogeneous, class-balanced, single-source, and accessible data. In order to improve the applicability of machine learning in helping the diagnosis and prognosis of neurodegenerative diseases, it is necessary to develop new data preprocessing methods and data representation learning methods.
This research topic aims to provide a diverse but complementary set of contributions to demonstrate new developments and applications that cover existing issues in data processing and analysis of neurodegenerative diseases data for helping diagnosis and prognosis. We would also like to accept successful applications of the new methods, including but not limited to data processing, analysis, and knowledge discovery of neurodegenerative diseases data for helping diagnosis and prognosis
| Original language | English |
|---|---|
| Journal | Frontiers in Computational Neuroscience |
| Publication status | In preparation - 19 Sept 2022 |