November 10, 2024

Medical Trend

Medical News and Medical Resources

Can Mild Cognitive Impairment Progress to Alzheimer’s?

Can Mild Cognitive Impairment Progress to Alzheimer’s? AI Model from Cambridge University Achieves 81.66% Prediction Accuracy.



Can Mild Cognitive Impairment Progress to Alzheimer’s? AI Model from Cambridge University Achieves 81.66% Prediction Accuracy.

Early diagnosis of Alzheimer’s disease (AD) has always been a significant challenge. Despite the unprecedented development of artificial intelligence in recent years, there is still a lack of sensitive tools for early identification of patients who may progress to AD, causing many to miss the optimal intervention window.

Recently, a team led by Zoe Kourtzi at the University of Cambridge developed an AI tool, the Predictive Prognostic Model (PPM), to address this issue. This model has been validated in multicenter cohorts in the UK and Singapore.

Overall, PPM can effectively predict whether patients with Mild Cognitive Impairment (MCI) will remain stable or progress to AD, with an accuracy of 81.66% and an area under the curve (AUC) of 0.84.

Additionally, the researchers developed a personalized predictive prognostic index based on PPM. Compared to standard clinical indicators like brain gray matter atrophy and cognitive test scores, this index more accurately predicts the risk of AD progression (HR=3.42).

The study was published in eClinicalMedicine.

 

Can Mild Cognitive Impairment Progress to Alzheimer's? AI Model from Cambridge University Achieves 81.66% Prediction Accuracy.

 

Training and Validation of the PPM Model

The researchers used a machine learning method based on a Generalized Matrix Learning Vector Quantization (GMLVQ) classification framework to train the PPM model. They input baseline MRI data (primarily temporal lobe gray matter density imaging to observe brain structure changes), Addenbrooke’s Cognitive Examination-Revised (ACE-R), and Mini-Mental State Examination (MMSE) results from 410 MCI patients in the Alzheimer’s Disease Neuroimaging Initiative (ADNI) cohort to train the PPM.

After training, PPM could categorize these patients into stable MCI (sMCI, individuals diagnosed with MCI over three years) and progressive MCI (pMCI, individuals who progressed to AD over three years).

The model was then validated using three datasets: the UK NHS Quantitative MRI dataset (QMIN-MC, n=272), the Memory Aging and Cognition Centre (MACC) dataset from the National University of Singapore (n=605), and another set of 609 participants from the ADNI cohort.

These datasets included various patient types to test PPM’s generalizability and clinical utility. The results showed that PPM effectively predicted whether MCI patients would remain stable or progress to AD, with an accuracy of 81.66% and an AUC of 0.84.

 

Expanding the Use of PPM from Diagnosis to Prognosis

To extend PPM’s utility from diagnosis to prognosis, researchers generated a prognostic index derived from PPM using a scalar projection method to evaluate individual cognitive function changes.

The findings revealed significant differences in this index among individuals with different cognitive states (normal cognition, MCI, and AD). These differences remained significant even after controlling for educational level, indicating the index’s capability to assess individual cognitive function changes.

To enhance PPM’s clinical utility, researchers developed a stratification method based on the PPM-derived prognostic index. They described three potential disease progression categories based on the rate of cognitive decline (i.e., future MMSE changes): stable (PPM index < 0), rapid progression (PPM index > 1), and slow progression (PPM index between 0 and 1).

Using six-year longitudinal data from the MACC cohort, the researchers stratified 387 participants with normal cognition and MCI (189 stable, 111 slow progression, and 87 rapid progression) based on the PPM-derived prognostic index and validated whether these stratifications accurately predicted the risk of future AD progression.

The results showed that stratification based on the PPM-derived prognostic index more accurately predicted the risk of AD conversion within three years compared to traditional clinical diagnostic standards.

Specifically, among those predicted as stable by PPM, only 0.5% converted to AD within three years; 18.9% of those predicted as slow progression converted to AD; and 41.4% of those predicted as rapid progression converted to AD.

In contrast, based on traditional clinical diagnosis, 3.2% of individuals with normal cognition, 11.8% of those with mild MCI, and 30.6% of those with severe MCI converted to AD within three years.

Moreover, multivariable Cox analysis showed that stratification based on the PPM-derived prognostic index predicted the risk of AD progression more accurately than traditional clinical diagnostic methods (HR=2.84).

Similarly, in terms of the PPM-derived prognostic index, multivariable Cox analysis demonstrated that it more accurately predicted the risk of AD progression than standard clinical indicators like gray matter atrophy MRI and MMSE scores (HR=3.42).

In conclusion, this study confirms that the AI tool based on PPM can accurately predict the risk of future AD progression, aiding early intervention and personalized treatment for AD. As the study’s corresponding author, Zoe Kourtzi, stated, “Our vision is to extend the application of our AI tool to help clinicians allocate the right diagnosis and treatment pathways to the right people at the right time.”

Can Mild Cognitive Impairment Progress to Alzheimer’s?

(source:internet, reference only)


Disclaimer of medicaltrend.org


Important Note: The information provided is for informational purposes only and should not be considered as medical advice.