P4‐588: END‐TO‐END 3D‐CONVOLUTIONAL NEURAL NETWORK FOR PREDICTING CONVERSION FROM MILD COGNITIVE IMPAIRMENT TO ALZHEIMER'S DEMENTIA
Bibliographic record
Abstract
Predicting conversion to Alzheimer's Dementia (AD) among Mild Cognitive Impairment (MCI) patients is invaluable for patient care, as well as in selection for clinical trials. This project utilized data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) to develop end-to-end 3D-Convolutional Neural Network (3D-CNN) models to classify subjects who did not progress to AD (i.e., stable MCI or sMCI) vs. subjects who did progress to AD (i.e., progressive MCI or pMCI). 470 sMCI and 293 pMCI patients’ skull-stripped baseline structural MRI (sMRI) scans were collected from ADNI and pre-processed by using crop, pad, bias field correction, and affine linear alignment with FMRIB Software Library (FSL). Three separate 3D-CNN approaches were developed and compared. First, ImageNet models were implemented, using the Residual Network50 with an augmentation by Squeeze-Excitation Network, i.e. SE_ResNet50. Second, transfer learning was implemented. The domain knowledge of classifying stable Normal Control (sNC) vs. stable Dementia of Alzheimer's Type (sDAT) was transferred to a reduced version of ResNet50, i.e., ResNet35. Third, a customized version of ResNet29 was implemented, i.e. cResNet29, where a residual block consisting of ResNet29 was modified to drop ungeneralizable features. The cResNet29 produced the highest test classification accuracy, 72.81% and the SE_ResNet50 recorded the second highest test accuracy, 70.01%. Based on these results, we confirmed that customizing ResNet performed better than augmenting with additional Network. Transfer learning resulted in the 67.54%. It might imply that the domain knowledge of classifying sNC vs. sDAT is not helpful in classifying sMCI vs. pMCI.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; both teacher heads agree on what is shown here.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".