Sign in Subscribe
Diseases

Deep Learning-Based Harmonization of Amyloid PET Improves Cognitive Decline Prediction in Non-Demented Elderly

Strategic Insights

— 2 min read

Researchers from the National University Health System in Singapore have made a significant breakthrough in the early detection and prediction of Alzheimer's disease using deep learning-based harmonization of amyloid PET images. The study, which involved the analysis of data from over 1,050 participants, found that the deep learning-based model was able to accurately predict cognitive decline in non-demented elderly individuals with a high degree of accuracy.

Key Takeaways:

  • The researchers developed a deep learning-based model to classify participants with a clinical diagnosis of Alzheimer's disease-dementia versus cognitively unimpaired individuals across different tracers from various cohorts, including the Alzheimer's Disease Neuroimaging Initiative (ADNI) and the Japanese ADNI.
  • The model output, known as deep learning-based probability of Alzheimer's disease-dementia (DL-ADprob), was evaluated for predicting cognitive decline in ADNI-MCI and Harvard Aging Brain Study (HABS)-CU participants using Cox regression and area under time-dependent receiver operating characteristics curve (tdAUC) at 4-year follow-up.
  • The study found that the deep learning-based model was independently prognostic in both ADNI-MCI ( Hazard Ratio (HR): 1.15, 95% CI: 1.06-1.25, p = 0.001) and HABS-CU (HR: 1.14, 95% CI: 1.05-1.24, p = 0.002) groups.
  • The DL-ADprob was also found to be independently prognostic in subgroup analyses performed in the ADNI-MCI group for conversion from amyloid-positive to AD and from amyloid negative to positive.
  • The intraclass correlation coefficient (ICC) of DL-ADprob between tracers was calculated in the Global Alzheimer's Association Interactive Network dataset (n = 155), demonstrating good between-tracer agreement.
  • The study concluded that deep learning-based harmonization of amyloid PET improves cognitive decline prediction in non-demented elderly, suggesting it could complement conventional amyloid PET measures.

Statistics:

  • The study involved the analysis of data from over 1,050 participants across various cohorts.
  • The deep learning-based model was evaluated for predicting cognitive decline in ADNI-MCI and HABS-CU participants using Cox regression and tdAUC at 4-year follow-up.
  • The DL-ADprob was found to be independently prognostic in both ADNI-MCI (HR: 1.15, 95% CI: 1.06-1.25, p = 0.001) and HABS-CU (HR: 1.14, 95% CI: 1.05-1.24, p = 0.002) groups.
  • The ICC of DL-ADprob between tracers was calculated in the Global Alzheimer's Association Interactive Network dataset (n = 155), demonstrating good between-tracer agreement (ICC: 0.85, 95% CI: 0.78-0.91).

Sources:

  • NewsRx. Reports from National University Health System Advance Knowledge in Alzheimer Disease (Deep learning-based amyloid PET harmonization to predict cognitive decline in non-demented elderly). Mental Health Weekly Digest. October 20, 2025; p 773.
  • Ngam, P. I., et al. (2024). Deep learning-based amyloid PET harmonization to predict cognitive decline in non-demented elderly. Radiology Advances, 1(2).
  • Choi, Y. S., et al. (2024). Deep learning-based harmonization of amyloid PET in predicting conversion from cognitively unimpaired to mild cognitive impairment and mild cognitive impairment to Alzheimer's disease. Alzheimer's Research & Therapy, 16(1), 1-12.

Read more