From dd9a68f67a8a9a4bdfde9e4bc608786096004b28 Mon Sep 17 00:00:00 2001
From: Fahad Almuqhim <f.almuqhim@gmail.com>
Date: Fri, 9 Jan 2026 10:12:15 -0500
Subject: [PATCH] Add harmonization and NeuroCLR softwares

---
 .../_posts/2026-01-09-ASD-AE-Harmonization.md | 30 +++++++++++++++++++
 software/_posts/2026-01-09-NeuroCLR.md        | 30 +++++++++++++++++++
 2 files changed, 60 insertions(+)
 create mode 100644 software/_posts/2026-01-09-ASD-AE-Harmonization.md
 create mode 100644 software/_posts/2026-01-09-NeuroCLR.md

diff --git a/software/_posts/2026-01-09-ASD-AE-Harmonization.md b/software/_posts/2026-01-09-ASD-AE-Harmonization.md
new file mode 100644
index 00000000..0b36dc16
--- /dev/null
+++ b/software/_posts/2026-01-09-ASD-AE-Harmonization.md
@@ -0,0 +1,30 @@
+---
+layout: project
+title: ASD-AE-Harmonization
+contributors: [Prof-S, falmuqhim]
+handle: AE-fMRI-Harmonization
+status: complete
+
+
+# Optional
+website: 
+grant:
+grant_url:
+image: /assets/images/software/open-source.png
+tagline: Autoencoder-based harmonization of multisite fMRI for robust autism classification
+tags: [software]
+
+# Data and code
+github: https://github.com/pcdslab/Autoencoder-fMRI-Harmonization
+neurovault:
+openneuro:
+figshare:
+figshare_names:
+osf: https://osf.io/d8253
+---
+{% include JB/setup %}
+
+We propose an autoencoder-based framework for harmonizing multisite functional magnetic resonance imaging (fMRI) data to improve the generalizability of machine learning models across imaging centers. Our approach leverages the non-linear representation learning capability of autoencoders to reduce site-specific variability while preserving biologically meaningful signal, without relying on additive or multiplicative statistical assumptions. Unlike traditional harmonization methods, our framework avoids data leakage by eliminating the need for access to data from all sites during model training. We design and evaluate multiple autoencoder variants, including AE, SAE, TAE, and DAE, and assess their effectiveness using the Autism Brain Imaging Data Exchange I (ABIDE-I) dataset comprising 1,035 subjects from 17 imaging centers. Experimental results using leave-one-site-out cross-validation demonstrate statistically significant improvements over baseline models (p < 0.01), with mean classification accuracy gains ranging from 3.41% to 5.04%. These findings highlight the effectiveness of autoencoder-based harmonization for reducing site effects, improving robustness on unseen sites, and enabling reliable downstream neuroimaging analyses.
+
+## Status 
+The method development for this work is complete
\ No newline at end of file
diff --git a/software/_posts/2026-01-09-NeuroCLR.md b/software/_posts/2026-01-09-NeuroCLR.md
new file mode 100644
index 00000000..31ccc581
--- /dev/null
+++ b/software/_posts/2026-01-09-NeuroCLR.md
@@ -0,0 +1,30 @@
+---
+layout: project
+title: NeuroCLR
+contributors: [Prof-S, falmuqhim]
+handle: NeuroCLR
+status: complete
+
+
+# Optional
+website: 
+grant:
+grant_url:
+image: /assets/images/software/open-source.png
+tagline: Self-supervised learning framework for disorder classification from fMRI data
+tags: [software]
+
+# Data and code
+github: https://github.com/pcdslab/NeuroCLR
+neurovault:
+openneuro:
+figshare:
+figshare_names:
+osf: https://osf.io/mf5xk
+---
+{% include JB/setup %}
+
+We propose NeuroCLR, a self-supervised contrastive learning framework for learning robust and generalizable neural representations directly from raw resting-state functional magnetic resonance imaging (rs-fMRI) data. Our approach leverages contrastive objectives, anatomically consistent sampling, and augmented views of unlabeled fMRI time series to extract invariant representations that are consistent across subjects, imaging sites, and diagnostic categories. Unlike supervised and disorder-specific SSL approaches, NeuroCLR is pre-trained in a disorder-agnostic manner, enabling effective transfer to downstream classification tasks with limited labeled data. We pre-train NeuroCLR on large-scale multisite neuroimaging data comprising more than 3,600 participants from 44 imaging centers and over 720,000 region-specific fMRI time series. The resulting pre-trained model is fine-tuned for multiple disorder-specific classification tasks and consistently outperforms both supervised deep learning models and SSL methods trained on single disorders. Extensive experiments demonstrate robust generalizability across sites, highlighting NeuroCLR’s ability to learn biologically meaningful and transferable representations from unlabeled fMRI data. These findings establish NeuroCLR as a scalable and reproducible self-supervised framework for multisite neuroimaging analysis and cross-disorder clinical modeling.
+
+## Status 
+The method development for this work is complete
\ No newline at end of file