Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by

Same author

Visual acuity in controls and patients measured with Maxwellian view and a 3 mm pupil: Examining potential effects of inherent and induced aberrations.

PloS one·2026

Same author

Reasoning in machine vision by learning fast and slow thinking.

Nature communications·2026

Same author

Microbiome-behavior coupling shapes infant adaptation to early maternal unpredictability.

Frontiers in microbiology·2026

Same author

InnerEye-HS: a disease-agnostic clinical tool for hippocampal segmentation.

Brain communications·2026

Same author

Microstructure imaging of prostate cancer by diffusion MRI.

Magma (New York, N.Y.)·2026

Same author

The Role of Dendritic Spines in Water Exchange Measurements With Diffusion MRI: Double Diffusion Encoding and Free-Waveform MRI.

NMR in biomedicine·2026

Search research articles

Related Experiment Video

Updated: Oct 19, 2025

Unilateral Lung Volume Analysis Using Micro-CT for Enhanced Assessment of Pulmonary Fibrosis in Preclinical Models

Unilateral Lung Volume Analysis Using Micro-CT for Enhanced Assessment of Pulmonary Fibrosis in Preclinical Models

Published on: June 20, 2025

Training data distribution significantly impacts the estimation of tissue microstructure with machine learning.

Noemi G Gyori^1,2, Marco Palombo¹, Christopher A Clark²

¹Centre for Medical Image Computing, Department of Computer Science, University College London, London, UK.

Magnetic Resonance in Medicine

|September 21, 2021

Summary

Supervised machine learning (ML) for quantitative MRI parameter mapping is sensitive to training data distribution. Careful selection is crucial, as uniform sampling can improve accuracy for atypical parameters but reduce precision for typical ones.

Keywords:

machine learning microstructure imaging model fitting quantitative MRI training data distribution

More Related Videos

Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model

Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model

Published on: August 16, 2020

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Published on: October 10, 2018

Related Experiment Videos

Last Updated: Oct 19, 2025

Unilateral Lung Volume Analysis Using Micro-CT for Enhanced Assessment of Pulmonary Fibrosis in Preclinical Models

Unilateral Lung Volume Analysis Using Micro-CT for Enhanced Assessment of Pulmonary Fibrosis in Preclinical Models

Published on: June 20, 2025

Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model

Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model

Published on: August 16, 2020

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Published on: October 10, 2018

Area of Science:

Biomedical Imaging
Machine Learning in Medical Physics

Background:

Quantitative MRI parameter mapping traditionally relies on model fitting.
Supervised machine learning (ML) offers a promising alternative for improved efficiency and accuracy.
The impact of training data characteristics on ML model performance requires thorough investigation.

Purpose of the Study:

To demonstrate and quantify the influence of varying training data distributions on the accuracy and precision of quantitative MRI parameter estimates derived from supervised ML.
To compare supervised ML fitting with traditional model fitting approaches.

Main Methods:

A two- and three-compartment biophysical model was fitted to in-vivo human brain and simulated diffusion MRI data.
Supervised ML models, including artificial neural networks and random forest regressors, were trained on datasets with diverse ground truth parameter distributions.
Parameter estimates from traditional fitting and various supervised ML approaches were compared using synthetic test data for accuracy and precision.

Main Results:

Training datasets mirroring healthy human data distributions yielded high precision but inaccurate estimates for atypical parameter combinations.
Uniform sampling across the entire plausible parameter space resulted in more accurate estimates for atypical parameters but potentially lower precision for typical ones.
Supervised ML parameter estimation accuracy is strongly dependent on the training data distribution.

Conclusions:

The distribution of training data significantly impacts the reliability of supervised ML for quantitative MRI parameter mapping.
High precision in ML-derived parameter maps can mask underlying bias, necessitating careful evaluation beyond visual inspection.
Understanding and optimizing training data distributions are critical for robust and accurate quantitative MRI analysis using supervised ML.