There are many active research projects accessing and applying shared ADNI data. Use the search above to find specific research focuses on the active ADNI investigations. This information is requested annually as a requirement for data access.
| Principal Investigator | |
| Principal Investigator's Name: | Mengqi Li |
| Institution: | Lanzhou University |
| Department: | School of Information Science & Engineering |
| Country: | |
| Proposed Analysis: | I am writing my undergraduate thesis titled "Eye Movement Feature Extraction and Its Application in Disease Diagnosis." Currently, I have the following situation: I want to use an existing dataset, for example, the one for autism, which has 48 features (all of which are float data type), 44 observations, and is a binary classification problem (autistic or normal). Previous researchers trained models using traditional machine learning algorithms such as random forest, decision tree, and SVM, but I noticed that no one has used deep learning techniques such as convolutional neural networks (CNN), although they are commonly used for image-related tasks. However, when I tried using CNN, the performance was not significantly better than the traditional machine learning models. During my last week's progress report, my supervisor asked me why I chose to use deep learning when other researchers did not. He explained that the reason might be due to the small size of the dataset, which limits the effectiveness of neural networks. Therefore, he suggested that I should consider expanding the dataset, and he was confident that there are methods available to deal with small medical datasets. However, I have searched online and found that most data augmentation techniques are used for audio, text, or image data, and I have not found any suitable methods for numerical data such as mine (although it is possible that I have not found all the available methods). If there are such methods, I can apply them to expand the dataset and then use it to train my deep learning models to see if the performance improves. If not, I will have no choice but to abandon this approach. Therefore, I am seeking a larger dataset to complete my undergraduate thesis, and I would be grateful if you could approve my request. |
| Additional Investigators |
