AI Insight
Researchers developed NITROGEN, a transformer-based machine learning model that predicts Alzheimer's disease diagnosis and severity without requiring imputation of missing clinical data. The model was trained on 7,858 brain scans and validated on two independent cohorts, demonstrating robust performance and well-calibrated uncertainty estimates across heterogeneous patient populations with incomplete records. The approach identified cortical thickness in the temporal pole, age, and APOE genotype as key predictive features while maintaining reliable predictions even when diagnostic information was partially absent.
Why it matters
This method addresses a critical barrier in clinical AI deployment by handling incomplete medical records without introducing the biases typical of data imputation techniques. The model's calibrated uncertainty quantification and cross-cohort reliability make it more suitable for real-world clinical settings where missing data is common and confident predictions on incomplete information could have serious consequences.
Understand the Science
arXiv:2607.11656v2 Announce Type: replace
Abstract: Accurate diagnostic classification and disease-severity prediction for Alzheimer’s disease are hampered by the incompleteness and heterogeneity of real-world clinical data. Left unaddressed, these barriers prevent reliable disease modelling and hinder effective clinical evaluation. Conventional imputation strategies introduce systematic bias, distort inter-feature relationships, and yield overconfident predictions, limitations especially consequential in diagnostic settings. Here, we propose NITROGEN, an imputation-free transformer that jointly models within-patient feature dependencies and between-patient relational structure through masked and intersample attention, enabling robust multimodal learning directly from partially observed records. We trained NITROGEN on ADNI (N=7858 scans), and evaluated it on two independent cohorts: OASIS-3 (N=2675 scans) and AIBL (N=1286 scans). Across cohorts and diagnostic and cognitive score prediction tasks, NITROGEN showed robust calibration and uncertainty quantification advantages over tree-based ensemble methods, while maintaining competitive discriminative performance. Cross-cohort and cross-method analyses identified cortical thickness in the temporal pole, age, and APOE genotype as important, though not individually sufficient, features for AD classification. We further introduced a modality-aware uncertainty adjustment that augments predictive uncertainty proportionally to the importance of absent modalities, enabling calibrated confidence when diagnostic information is unavailable. Together, our results show that imputation-free attention learning preserved meaningful discrimination under cohort shift, revealing expected degradation on more distributionally different cohorts, and demonstrate that evaluating models along calibration, interpretability, and cross-cohort reliability, not accuracy alone, is essential for clinical deployment.