Brain MR image processing with deep neural networks | |
| Author | Praewphan Tocharoenkul |
| Call Number | AIT RSPR no.DSAI-23-04 |
| Subject(s) | Neural networks (Computer science) Brain--Diseases--Data processing Alzheimer's Disease |
| Note | A research study submitted in partial fulfillment of the requirements for the degree of Master of Engineering in Data Science and Artificial Intelligence, School of Environment, Resources and Development |
| Publisher | Asian Institute of Technology |
| Series Statement | Research studies project report ; no. DSAI-23-04 |
| Abstract | At present, there is no cure for Alzheimer’s disease. However, there are treatment options for patients to have a better quality of life. Treatment makes use of acetyl cholinesterase inhibitorsto reduce destruction of the structures of the brain. It would be very helpful if we can diagnose Alzheimer’s disease early and accurately. In the literature, there are currently no models able to automatically classify abnormalities related to Alzheimer’s disease in terms of severity level. I aim to develop a model able to classify each lobe of the brain in term of severity. To achieve the aim, I performed three steps to developed models. Frist, I built a dataset for training atrophy severity level classification models based on GCA assessment, then I segmented the dataset according to the lobes of the brain. Finally, I aim to build a model suitable for classifying each lobe of the brain in terms of severity. In the experiments, the models can classify the severity level for each lobe of brain. The overall model obtained a test accuracy of 62.75%, broken down into the four classes with frontal accuracy of 68.0%, parietal accuracy of 54.0%, a temporal accuracy of 62.0%, and occipital accuracy of 67.0%. There are some limitations to the experiments. First, I focused mainly on the axial view, which was recommended by Dr. Dittapong in the initial studying stage. Because the dataset is complicated and sensitive to label accurately, there may be discrepancies in the dataset. Due to the project time limitations, I was unable to obtain substantial medical a training and knowledge. To segment the dataset into each lobe of the brain, I picked a range of specific slices for each lobe of the brain manually. Within the range I chose, there are some regions that are not specific for the lobe, which may affect the accuracy of the model. From the model in this thesis, I further understand brain anatomy, especially the structure of sulci and gyri, specific parts of the brain, limitations of the dataset, and how to build the model to classify the severity level for each lobe of brain. In this thesis, I developed an initial version of such a mode. While the result is not yet to be ready for clinical utilization, it has laid some foundations for future research. |
| Year | 2023 |
| Corresponding Series Added Entry | Asian Institute of Technology. Research studies project report ; no. DSAI-23-04 |
| Type | Research Study Project Report (RSPR) |
| School | School of Engineering and Technology |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Data Science and Artificial Intelligence (DSAI) |
| Chairperson(s) | Dailey, Matthew N.; |
| Examination Committee(s) | Chaklam Silpasuwanchai;Dittapong Songsaeng; |
| Scholarship Donor(s) | His Majesty the King’s Scholarships; |
| Degree | Research Studies Project Report (M. Eng.) - Asian Institute of Technology, 2023 |