Evaluating TabPFN for Mild Cognitive Impairment to Alzheimer's Disease Conversion in Data Limited Settings

TL;DR

This study evaluates the effectiveness of TabPFN, a foundation model, in predicting 3-year conversion from Mild Cognitive Impairment to Alzheimer's Disease using limited data, outperforming traditional methods especially in small sample settings.

Contribution

The paper demonstrates that TabPFN outperforms traditional machine learning models in predicting Alzheimer's conversion with limited data, highlighting its potential in data-scarce medical scenarios.

Findings

TabPFN achieved an AUC of 0.892, outperforming LightGBM's 0.860.

TabPFN maintained strong performance with as few as 50 training samples.

Traditional models struggled with small training datasets, unlike TabPFN.

Abstract

Accurate prediction of conversion from Mild Cognitive Impairment (MCI) to Alzheimers Diseases (AD) is essential for early intervention, however, developing reliable conversion predictive models is difficult to develop due to limited longitudinal data availability We evaluate TabPFN (Tabular Pre-Trained Foundation Network) against traditional machine learning methods for predicting 3 year MCI to AD conversion using the TADPOLE dataset derived from ADNI. Using multimodal biomarker features extracted from demographics, APOE4, MRI volumes, CSF markers, and PET imaging, we conducted an experimental comparison across varying training set sizes (N=50 to 1000) and models including XGBoost, Random Forest, LightGBM, and Logistic Regression. TabPFN achieved one the highest performance (AUC=0.892), outperforming LightGBM (AUC=0.860) and demonstrating advantages in low data settings. At N=50 training samples, TabPFN maintained strong AUC while the traditional machine learning models struggles at small training samples. These findings demonstrate that foundation models are promising for disease prediction in data limited scenarios, such as Alzheimers diseases.