Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Head Motion Degrades Machine Learning Classification of Alzheimer's Disease from Positron Emission Tomography

Eléonore V. Lieffrig, Takuya Toyonaga, Jiazhen Zhang, John A. Onofrey

TL;DR

This paper addresses how head motion degrades PET-based Alzheimer’s disease classification by comparing motion-corrected (MC) and non-motion-corrected (NMC) reconstructions using two clinically relevant tracers, $^{11}$C-UCB-J and $^{18}$F-FDG. The authors propose a PET-only binary CN/AD classifier, leveraging ResNet10-based feature extraction and SVM classification, with various pretraining strategies on MedicalNet; zero-shot transfer generally yields the best cross-tracer performance. Key findings show AUROCs of about $0.94$–$0.96$ on MC data, but notable performance drops when motion correction is absent (up to $10 ext{%}$ accuracy loss for FDG and $5 ext{%}$ for UCB-J; recall can drop from $1$ to lower values). The results underscore the critical need for portable motion-correction methods to unlock the full diagnostic potential of PET-based ML in AD, and suggest that MedicalNet pretraining plus SVM on features provides a robust approach across tracers.

Abstract

Brain positron emission tomography (PET) imaging is broadly used in research and clinical routines to study, diagnose, and stage Alzheimer's disease (AD). However, its potential cannot be fully exploited yet due to the lack of portable motion correction solutions, especially in clinical settings. Head motion during data acquisition has indeed been shown to degrade image quality and induces tracer uptake quantification error. In this study, we demonstrate that it also biases machine learning-based AD classification. We start by proposing a binary classification algorithm solely based on PET images. We find that it reaches a high accuracy in classifying motion corrected images into cognitive normal or AD. We demonstrate that the classification accuracy substantially decreases when images lack motion correction, thereby limiting the algorithm's effectiveness and biasing image interpretation. We validate these findings in cohorts of 128 $^{11}$C-UCB-J and 173 $^{18}$F-FDG scans, two tracers highly relevant to the study of AD. Classification accuracies decreased by 10% and 5% on 20 $^{18}$F-FDG and 20 $^{11}$C-UCB-J testing cases, respectively. Our findings underscore the critical need for efficient motion correction methods to make the most of the diagnostic capabilities of PET-based machine learning.

Head Motion Degrades Machine Learning Classification of Alzheimer's Disease from Positron Emission Tomography

TL;DR

This paper addresses how head motion degrades PET-based Alzheimer’s disease classification by comparing motion-corrected (MC) and non-motion-corrected (NMC) reconstructions using two clinically relevant tracers, C-UCB-J and F-FDG. The authors propose a PET-only binary CN/AD classifier, leveraging ResNet10-based feature extraction and SVM classification, with various pretraining strategies on MedicalNet; zero-shot transfer generally yields the best cross-tracer performance. Key findings show AUROCs of about – on MC data, but notable performance drops when motion correction is absent (up to accuracy loss for FDG and for UCB-J; recall can drop from to lower values). The results underscore the critical need for portable motion-correction methods to unlock the full diagnostic potential of PET-based ML in AD, and suggest that MedicalNet pretraining plus SVM on features provides a robust approach across tracers.

Abstract

Brain positron emission tomography (PET) imaging is broadly used in research and clinical routines to study, diagnose, and stage Alzheimer's disease (AD). However, its potential cannot be fully exploited yet due to the lack of portable motion correction solutions, especially in clinical settings. Head motion during data acquisition has indeed been shown to degrade image quality and induces tracer uptake quantification error. In this study, we demonstrate that it also biases machine learning-based AD classification. We start by proposing a binary classification algorithm solely based on PET images. We find that it reaches a high accuracy in classifying motion corrected images into cognitive normal or AD. We demonstrate that the classification accuracy substantially decreases when images lack motion correction, thereby limiting the algorithm's effectiveness and biasing image interpretation. We validate these findings in cohorts of 128 C-UCB-J and 173 F-FDG scans, two tracers highly relevant to the study of AD. Classification accuracies decreased by 10% and 5% on 20 F-FDG and 20 C-UCB-J testing cases, respectively. Our findings underscore the critical need for efficient motion correction methods to make the most of the diagnostic capabilities of PET-based machine learning.
Paper Structure (10 sections, 1 figure, 3 tables)

This paper contains 10 sections, 1 figure, 3 tables.

Table of Contents

  1. Introduction
  2. Methods
  3. PET Image Reconstruction
  4. AD Classification Algorithm
  5. Experiments and Results
  6. Experimental Setup
  7. Results
  8. Discussion and Conclusion
  9. Compliance with ethical standards
  10. Acknowledgments

Figures (1)

  • Figure 2: AUROC curves comparing the best model performances on testing subjects with motion correction (MC) and without motion correction (NMC).