NeuroImage: Clinical (Jan 2023)
Validation of deep learning techniques for quality augmentation in diffusion MRI for clinical studies
- Santiago Aja-Fernández,
- Carmen Martín-Martín,
- Álvaro Planchuelo-Gómez,
- Abrar Faiyaz,
- Md Nasir Uddin,
- Giovanni Schifitto,
- Abhishek Tiwari,
- Saurabh J. Shigwan,
- Rajeev Kumar Singh,
- Tianshu Zheng,
- Zuozhen Cao,
- Dan Wu,
- Stefano B. Blumberg,
- Snigdha Sen,
- Tobias Goodwin-Allcock,
- Paddy J. Slator,
- Mehmet Yigit Avci,
- Zihan Li,
- Berkin Bilgic,
- Qiyuan Tian,
- Xinyi Wang,
- Zihao Tang,
- Mariano Cabezas,
- Amelie Rauland,
- Dorit Merhof,
- Renata Manzano Maria,
- Vinícius Paraníba Campos,
- Tales Santini,
- Marcelo Andrade da Costa Vieira,
- SeyyedKazem HashemizadehKolowri,
- Edward DiBella,
- Chenxu Peng,
- Zhimin Shen,
- Zan Chen,
- Irfan Ullah,
- Merry Mani,
- Hesam Abdolmotalleby,
- Samuel Eckstrom,
- Steven H. Baete,
- Patryk Filipiak,
- Tanxin Dong,
- Qiuyun Fan,
- Rodrigo de Luis-García,
- Antonio Tristán-Vega,
- Tomasz Pieciak
Affiliations
- Santiago Aja-Fernández
- Laboratorio de Procesado de Imagen (LPI), ETSI Telecomunicación, Universidad de Valladolid, Spain; Corresponding author at: ETSI Telecomunicación, Universidad de Valladolid, Spain.
- Carmen Martín-Martín
- Laboratorio de Procesado de Imagen (LPI), ETSI Telecomunicación, Universidad de Valladolid, Spain
- Álvaro Planchuelo-Gómez
- Laboratorio de Procesado de Imagen (LPI), ETSI Telecomunicación, Universidad de Valladolid, Spain; Cardiff University Brain Research Imaging Centre (CUBRIC), School of Psychology, Cardiff University, Cardiff, UK
- Abrar Faiyaz
- University of Rochester, USA
- Md Nasir Uddin
- University of Rochester, USA
- Giovanni Schifitto
- University of Rochester, USA
- Abhishek Tiwari
- Shiv Nadar Institution of Eminence, India
- Saurabh J. Shigwan
- Shiv Nadar Institution of Eminence, India
- Rajeev Kumar Singh
- Shiv Nadar Institution of Eminence, India
- Tianshu Zheng
- Zhejiang University, China
- Zuozhen Cao
- Zhejiang University, China
- Dan Wu
- Zhejiang University, China
- Stefano B. Blumberg
- University College London, UK
- Snigdha Sen
- University College London, UK
- Tobias Goodwin-Allcock
- University College London, UK
- Paddy J. Slator
- University College London, UK
- Mehmet Yigit Avci
- Athinoula A. Martinos Center for Biomedical Imaging, USA
- Zihan Li
- Athinoula A. Martinos Center for Biomedical Imaging, USA
- Berkin Bilgic
- Athinoula A. Martinos Center for Biomedical Imaging, USA
- Qiyuan Tian
- Athinoula A. Martinos Center for Biomedical Imaging, USA
- Xinyi Wang
- University of Sydney, Australia
- Zihao Tang
- University of Sydney, Australia
- Mariano Cabezas
- University of Sydney, Australia
- Amelie Rauland
- RWTH Aachen University, Germany
- Dorit Merhof
- University of Regensburg, Germany
- Renata Manzano Maria
- University of Sao Paulo, Brazil
- Vinícius Paraníba Campos
- University of Sao Paulo, Brazil
- Tales Santini
- Western University, Canada
- Marcelo Andrade da Costa Vieira
- University of Sao Paulo, Brazil
- SeyyedKazem HashemizadehKolowri
- University of Utah, USA
- Edward DiBella
- University of Utah, USA
- Chenxu Peng
- Zhejiang University of Technology, China
- Zhimin Shen
- Zhejiang University of Technology, China
- Zan Chen
- Zhejiang University of Technology, China
- Irfan Ullah
- University of Iowa, USA
- Merry Mani
- University of Iowa, USA
- Hesam Abdolmotalleby
- University of Iowa, USA
- Samuel Eckstrom
- New York University, USA
- Steven H. Baete
- New York University, USA
- Patryk Filipiak
- New York University, USA
- Tanxin Dong
- Tianjin University, China
- Qiuyun Fan
- Tianjin University, China
- Rodrigo de Luis-García
- Laboratorio de Procesado de Imagen (LPI), ETSI Telecomunicación, Universidad de Valladolid, Spain
- Antonio Tristán-Vega
- Laboratorio de Procesado de Imagen (LPI), ETSI Telecomunicación, Universidad de Valladolid, Spain
- Tomasz Pieciak
- Laboratorio de Procesado de Imagen (LPI), ETSI Telecomunicación, Universidad de Valladolid, Spain
- Journal volume & issue
-
Vol. 39
p. 103483
Abstract
The objective of this study is to evaluate the efficacy of deep learning (DL) techniques in improving the quality of diffusion MRI (dMRI) data in clinical applications. The study aims to determine whether the use of artificial intelligence (AI) methods in medical images may result in the loss of critical clinical information and/or the appearance of false information. To assess this, the focus was on the angular resolution of dMRI and a clinical trial was conducted on migraine, specifically between episodic and chronic migraine patients. The number of gradient directions had an impact on white matter analysis results, with statistically significant differences between groups being drastically reduced when using 21 gradient directions instead of the original 61. Fourteen teams from different institutions were tasked to use DL to enhance three diffusion metrics (FA, AD and MD) calculated from data acquired with 21 gradient directions and a b-value of 1000 s/mm2. The goal was to produce results that were comparable to those calculated from 61 gradient directions. The results were evaluated using both standard image quality metrics and Tract-Based Spatial Statistics (TBSS) to compare episodic and chronic migraine patients. The study results suggest that while most DL techniques improved the ability to detect statistical differences between groups, they also led to an increase in false positive. The results showed that there was a constant growth rate of false positives linearly proportional to the new true positives, which highlights the risk of generalization of AI-based tasks when assessing diverse clinical cohorts and training using data from a single group. The methods also showed divergent performance when replicating the original distribution of the data and some exhibited significant bias. In conclusion, extreme caution should be exercised when using AI methods for harmonization or synthesis in clinical studies when processing heterogeneous data in clinical studies, as important information may be altered, even when global metrics such as structural similarity or peak signal-to-noise ratio appear to suggest otherwise.
