Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Hierarchical Information Aggregation for Incomplete Multimodal Alzheimer's Disease Diagnosis
Authors: Chengliang Liu, Que Yuanxi, Qihao Xu, Yabo Liu, Jie Wen, Jinghua Wang, Xiaoling Luo
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 4 Experiments 4.1 Experimental settings 4.2 Experimental Analysis 4.3 Modality Imbalance Study 4.4 Ablation Study Comprehensive experimental evaluations confirm that our HAD consistently achieves significant performance advantages under various modality-missing cases. |
| Researcher Affiliation | Academia | 1College of Computer Science and Software Engineering, Shenzhen University 2Laboratory for Artificial Intelligence in Design, The Hong Kong Polytechnic University 3School of Computer Science and Technology, Harbin Institute of Technology, Shenzhen 4College of Artificial Intelligence, Ocean University of China |
| Pseudocode | No | The paper describes methods using equations and descriptive text, but no explicit pseudocode blocks or algorithms are present. |
| Open Source Code | Yes | However, we have provided the complete code utilized in our analyses within the supplementary materials. Clear instructions are included to ensure reviewers and interested researchers can faithfully reproduce our main experimental results upon obtaining authorized access to the ADNI dataset. |
| Open Datasets | Yes | The data utilized in this study is collected from the Alzheimer s Disease Neuroimaging Initiative (ADNI), a publicly available database designed to facilitate research into biomarkers and clinical trials for AD. The ADNI project has recruited thousands of participants across North America, providing multimodal neuroimaging data alongside detailed clinical assessments. Specifically, we select baseline T1-weighted structural MRI and paired 18F-AV45 PET images as bi-modal brain imaging; we collect the values of biomarkers, such as amyloid β-protein (Aβ), Tau, and p-Tau, as CSF data; and 29 clinical cognitive examination scores as the CAD. All data is from four ADNI subsets: ADNI-1, ADNI-2, ADNI-3, and ADNI-GO. |
| Dataset Splits | Yes | Then, all subjects are divided into 5 subsets to facilitate the 5-fold cross-validation. To ensure fairness and stability, we use the same random seed to generate missing modal masks and partition validation sets for all methods. |
| Hardware Specification | No | The paper does not explicitly describe the specific hardware (e.g., GPU/CPU models, memory) used for running its experiments in the main text or the provided appendix. |
| Software Dependencies | No | For preprocessing, PET images are first aligned to their corresponding MRI scans. Subsequently, both MRI and PET images are spatially normalized to the standard Montreal Neurological Institute (MNI) space using Statistical Parametric Mapping (SPM) [33]. Intensity normalization and Gaussian smoothing are also applied to PET images to reduce image noise and standardize intensity values. Finally, skull-stripping procedure is conducted on both MRI and PET images using Free Surfer [34] to remove non-brain tissues and further enhance data quality for subsequent analysis. |
| Experiment Setup | No | The paper discusses the optimization objective (L = Lce + λLintra + γLinter) and describes the modality imbalance study at the "20th training epoch" for visualizations, but it does not provide specific hyperparameter values like learning rate, batch size, or detailed optimizer configurations in the main text. |