Predicting Amyloid: AI for Early Detection

10 février 2025

#Alzheimer’s #MachineLearning #Neurology #Screening #AmyloidPlaque The accumulation of amyloid plaques in the brain is a key pathological feature of Alzheimer’s disease. These plaques often appear years, even decades, before cognitive symptoms emerge. Their accumulation is associated with an increased risk of cognitive decline, making them a priority target for screening and early intervention. Ho

#Alzheimer’s #MachineLearning #Neurology #Screening #AmyloidPlaque

The accumulation of amyloid plaques in the brain is a key pathological feature of Alzheimer’s disease. These plaques often appear years, even decades, before cognitive symptoms emerge. Their accumulation is associated with an increased risk of cognitive decline, making them a priority target for screening and early intervention. However, current detection tools, such as lumbar puncture for biomarker measurement in cerebrospinal fluid (CSF) or positron emission tomography (PET) imaging, are invasive, expensive, and not widely accessible, limiting their use in asymptomatic populations.

Faced with these challenges, this study aimed to develop predictive models to detect cerebral amyloid positivity using non-invasive data. These data include demographic characteristics, cognitive tests, and accessible blood biomarkers. The model developed must be effective, affordable, and easily integrated into clinical practice. It has the potential to revolutionize cerebral amyloidosis screening by enabling early identification of at-risk individuals, paving the way for targeted interventions before symptom onset.

Read next: Alzheimer’s Disease: The Amyloid Hypothesis Is Not the Only Factor

Predicting Alzheimer’s Without a Lumbar Puncture – Is It Possible?

Using the French MEMENTO cohort, which includes 853 non-demented participants, researchers worked on developing predictive models capable of identifying amyloid-positive patients. These models were built by integrating various predictors: demographic data (age, sex), cognitive assessments, blood biomarkers (Aβ42/40 and P-tau181 ratios), and ApoE4 status, which is associated with Alzheimer's disease risk. External validation was conducted with the Amsterdam Dementia Cohort (ADC) to assess the models' generalizability. Predictive performance was measured using indicators such as the Area Under the Curve (AUC) and calibration curves (accuracy and robustness).

The model incorporating demographic, cognitive, and blood biomarker data demonstrated superior performance, achieving an AUC of 0.82 in the MEMENTO cohort and 0.90 in the ADC cohort. Compared to models based solely on demographic and cognitive data, this combined approach significantly improved accuracy. In contrast, adding brain imaging data, such as hippocampal atrophy and lobar microbleeds, did not enhance predictive capability, highlighting the importance of blood biomarkers in this context.

These findings underscore the potential of non-invasive biomarkers as key elements in the early detection of amyloid-positive patients, paving the way for more accessible and clinically relevant predictive tools.

Read next: A New Immunological Test for Alzheimer’s Disease

Towards a Simpler and More Accessible Diagnosis

Alzheimer’s disease is characterized by the progressive accumulation of amyloid plaques in the brain, a process that begins long before cognitive symptoms appear. Detecting this amyloid signature at an early stage is crucial for intervention before irreversible damage occurs. However, current detection methods have significant limitations, reducing their usefulness for large-scale systematic screening.

To offer a less invasive and more accessible alternative, this study aimed to develop a predictive model combining clinical data and blood biomarkers. Using the MEMENTO cohort for development and the ADC cohort for validation, the model sought to evaluate the accuracy and robustness of amyloid positivity predictions.

Results demonstrated that integrating blood biomarkers with demographic and cognitive data significantly improved predictive accuracy. With high AUC scores in both cohorts, the model outperformed traditional approaches based solely on clinical criteria. These findings represent a major advancement toward democratizing amyloid positivity detection.

Further validation on more diverse cohorts and under real-world conditions is still needed to ensure the model’s reliability. Additionally, access to blood biomarkers could be a limiting factor in certain regions, and the impact of early screening on treatment outcomes remains to be proven. Future research should refine these models by integrating genetic and environmental data and assessing their effectiveness in clinical trials. A broader adoption of these tools could transform Alzheimer’s disease diagnosis, making screening both earlier and more accessible.

Read next: Oncology: When AI Takes Control…

Reserved for healthcare professionals, free registration.

Scientific reference

Le Scouarnec, L., et al. (2024). Development and assessment of algorithms for predicting brain amyloid positivity in a population without dementia. Alzheimer's Research & Therapy, 16(1), 219

Voir la source

Autres articles récents

Does Nicotinamide Really Prevent Skin Cancer?

1 juillet 2026

Exercising in Extreme Heat: How Can Performance Be Preserved While Reducing Health Risks?

29 juin 2026

Oropouche Virus: A New Threat to Monitor During Pregnancy?

26 juin 2026

Toute la revue de presse