Deep clustering with convolution autoencoders and edge detection based classifcation and visualization of Alzheimer’s disease
| Autor: | A. Nivethitha, D. Baskar, L. Murali, T. Anandaselvakarthik |
|---|---|
| Rok vydání: | 2022 |
| Předmět: | |
| Zdroj: | International journal of health sciences. :10288-10302 |
| ISSN: | 2550-696X 2550-6978 |
| DOI: | 10.53730/ijhs.v6ns2.7751 |
| Popis: | Although there is no treatment for ADs (Alzheimer's Diseases), accurate and early diagnosis is critical for both patients and caregivers, and it will become much more vital if disease-modifying medicines are available to prevent, cure, or even halt the illness's course. One of the most active study topics in the medical industry in recent years has been the categorization of ADS using deep learning algorithms. However, most existing approaches are unable to utilise all spatial information, and so lose inter-slice correlation. To avoid this issues in recent works introduces CAEs (convolution auto encoders) based unsupervised learning for classifying ADs from NCs (normal controls), and supervised transfer learning is applied to solve classifications of Ads into pMCIs (progressive mild cognitive impairments) and sMCIs (stable mild cognitive impairments). A gradient-based visualisation technique that approximates the geographical effect of CNNs (Convolution Neural Networks) decisions were used to determine the most relevant biomarkers connected to ADs and pCMIs. Despite the fact that DLTs (deep learning techniques) perform well, finding optimal performing network structures for certain applications is not easy since it is frequently unclear how network structure affects network accuracies. This can be solved through hyper parameter tuning of DLTs. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|