[Paper Review] A Review on Image- and Network-based Brain Data Analysis Techniques for Alzheimer's Disease Diagnosis Reveals a Gap in Developing Predictive Methods for Prognosis

Mayssa Soussia, Islem Rekik|arXiv (Cornell University)|Aug 6, 2018

Dementia and Cognitive Impairment Research30 references3 citations

TL;DR

This review analyzes image- and network-based brain data techniques for Alzheimer’s disease (AD) diagnosis from MICCAI 2010–2016, revealing a critical gap in predictive modeling for MCI progression. While classification methods for distinguishing AD, MCI, and normal controls show high accuracy (up to 96.7%), few studies develop models to predict long-term MCI outcomes—such as conversion to AD, stability, or reversal—highlighting a major unmet need in early prognosis.

ABSTRACT

Unveiling pathological brain changes associated with Alzheimer's disease (AD) is a challenging task especially that people do not show symptoms of dementia until it is late. Over the past years, neuroimaging techniques paved the way for computer-based diagnosis and prognosis to facilitate the automation of medical decision support and help clinicians identify cognitively intact subjects that are at high-risk of developing AD. As a progressive neurodegenerative disorder, researchers investigated how AD affects the brain using different approaches: 1) image-based methods where mainly neuroimaging modalities are used to provide early AD biomarkers, and 2) network-based methods which focus on functional and structural brain connectivities to give insights into how AD alters brain wiring. In this study, we reviewed neuroimaging-based technical methods developed for AD and mild-cognitive impairment (MCI) classification and prediction tasks, selected by screening all MICCAI proceedings published between 2010 and 2016. We included papers that fit into image-based or network-based categories. The majority of papers focused on classifying MCI vs. AD brain states, which has enabled the discovery of discriminative or altered brain regions and connections. However, very few works aimed to predict MCI progression based on early neuroimaging-based observations. Despite the high importance of reliably identifying which early MCI patient will convert to AD, remain stable or reverse to normal over months/years, predictive models are still lagging behind.

Motivation & Objective

• To evaluate the state of the art in image- and network-based brain data analysis for Alzheimer’s disease (AD) diagnosis using MICCAI 2010–2016 proceedings.
• To identify the limitations in current methods, particularly the lack of predictive models for MCI progression to AD, stability, or reversal.
• To highlight the underuse of morphological brain networks and shape-based analysis in predictive frameworks despite their potential for early diagnosis.
• To emphasize the need for reproducible, generalizable, and publicly shared machine learning methods for prognosis in dementia research.
• To advocate for integrating advanced network and shape analysis with machine learning to improve long-term prediction of MCI outcomes.

Proposed method

• Systematic review of 28 papers from MICCAI 2010–2016, selected via keyword-based search on AD, MCI, neuroimaging, network, classification, and prediction.
• Categorization of studies into image-based methods (using structural, functional, and diffusion MRI) and network-based methods (focusing on functional and structural brain connectivity).
• Evaluation of classification performance using metrics such as accuracy, AUC, and F1-score across NC vs. AD, NC vs. MCI, and cMCI vs. sMCI tasks.
• Analysis of predictive models focusing on MCI conversion to AD using baseline MRI, with evaluation at 12- and 18-month follow-ups.
• Identification of missing methodological elements: absence of morphological brain networks and full trajectory prediction models for MCI evolution.
• Comparison of results across studies, noting inconsistencies due to differing datasets, baselines, and evaluation protocols.

Experimental results

Research questions

• RQ1 What are the dominant image- and network-based methods used for AD and MCI classification in neuroimaging studies from 2010 to 2016?
• RQ2 How effective are current methods in predicting MCI progression to AD, stability, or reversal over time?
• RQ3 Why is there a lack of predictive models for MCI outcomes despite high-performing classification techniques?
• RQ4 To what extent are morphological features such as cortical thickness and brain shape changes leveraged in existing predictive frameworks?
• RQ5 What are the key barriers to reproducibility and generalizability in current brain data analysis methods for AD prognosis?

Key findings

• The majority of reviewed studies focused on classification tasks, with image-based methods achieving classification accuracy up to 96.7% for distinguishing cMCI from sMCI.
• Network-based methods showed strong performance in NC/MCI classification, with accuracy reaching 96.59%, but were underutilized for prognosis.
• Only three studies—[33], [34], and [32]—reported predictive models for MCI conversion, with prediction accuracy of 76.53% at 18 months and 79.83% at 12 months before conversion.
• Despite evidence that morphological features like cortical thickness are altered in AD, no study in the review used morphological brain networks for prediction.
• No method was identified that predicts the full trajectory of brain shape changes across MCI progression, stability, or reversal to normal.
• The lack of publicly shared models and inconsistent evaluation protocols hinder reproducibility and generalizability across datasets.