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 [1].
NeuroTree: Hierarchical Functional Brain Pathway Decoding for Mental Health Disorders
Authors: Jun-En Ding, Dongsheng Luo, Chenwei Wu, Feng Liu
ICML 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our empirical evaluations demonstrate that Neuro Tree achieves state-of-the-art performance across two distinct mental disorder datasets. It provides valuable insights into age-related deterioration patterns, elucidating their underlying neural mechanisms. The code and datasets are available at https://github.com/Ding1119/Neuro Tree. |
| Researcher Affiliation | Academia | 1Department of Systems Engineering, Stevens Institute of Technology, New Jersey, USA 2Department of Computing and Information Sciences, Florida International University, USA 3Department of Electrical Engineering and Computer Science, University of Michigan, USA. Correspondence to: Feng Liu <EMAIL>. |
| Pseudocode | Yes | Algorithm 1 Hierarchical Tree Trunk Extraction with Weighted Tree Path Learning |
| Open Source Code | Yes | The code and datasets are available at https://github.com/Ding1119/Neuro Tree. |
| Open Datasets | Yes | We validated two publicly available f MRI datasets one focusing on cannabis use disorder and the other on schizophrenia. ①Cannabis (Kulkarni et al., 2023): The cannabis dataset comprises f MRI data from two distinct sources. ②COBRE (Calhoun et al., 2012): The Center for Biomedical Research Excellence (COBRE) dataset includes resting-state f MRI data collected from healthy controls and individuals diagnosed with schizophrenia. The code and datasets are available at https://github.com/Ding1119/Neuro Tree. |
| Dataset Splits | Yes | The comparative analysis was conducted using 5-fold cross-validation on two distinct datasets, benchmarking NEUROTREE against both baseline models and SOTA approaches. |
| Hardware Specification | Yes | Table 4. Experiment Environment Details Component Details GPU NVIDIA Ge Force RTX 3070 Python Version 3.7.6 Numpy Version 1.21.6 Pandas Version 1.3.5 Torch Version 1.13.1 Operating System Linux Processor x86 64 Architecture 64-bit Logical CPU Cores 32 Physical CPU Cores 24 Total Memory (GB) 62.44 |
| Software Dependencies | Yes | Table 4. Experiment Environment Details Component Details GPU NVIDIA Ge Force RTX 3070 Python Version 3.7.6 Numpy Version 1.21.6 Pandas Version 1.3.5 Torch Version 1.13.1 Operating System Linux Processor x86 64 Architecture 64-bit Logical CPU Cores 32 Physical CPU Cores 24 Total Memory (GB) 62.44 |
| Experiment Setup | Yes | Parameters Setting In this study, two datasets were trained with a batch size of 16 for 100 epochs using a learning rate of 0.001. Table 3. Training Parameters. Notation Meaning Value ρ Scale parameter 0.5 T Number of time segment 2 λ k-hop connectivity balance between dynamic FC matrices Ad(t) (0,1) β Learnable age-modulated parameter (0,1) Γ Learnable graph weight parameter W (l) k Learnable weight matrix in lth layer - |