A Comparative Study on Diagnosing Parkinson’s Disease Using Two Novel Neuromelanin MRI Sequences

266 

Abstract Submission 

Ming-Chih Kuo^1,2, Yao-Chia Shih¹, Ru-Jen Lin³, Kai-Hsiang Chen³, Yuh-Fen Wei⁴, Yen-Chung Hsiue⁴, Hui-Yu Yang³, Xue-Zhe Lu⁵, Joshua Oon Soo Goh^6,7,8,9

¹Graduate Institute of Medicine, Yuan Ze University, Taoyuan, Taiwan, ²Department of Medical Imaging, National Taiwan University Cancer Center, Taipei, Taiwan, ³Department of Neurology, National Taiwan University Hospital Hsin-Chu Branch, Hsinchu City, Taiwan, ⁴Department of Medical Imaging, National Taiwan University Hospital Hsin-Chu Branch, Hsinchu City, Taiwan, ⁵Siemens Healthineers Limited, Taipei, Taipei, ⁶Graduate Institute of Brain and Mind Sciences, College of Medicine, National Taiwan University, Taipei, Taiwan, ⁷Department of Psychology, National Taiwan University, Taipei, Taiwan, ⁸Neurobiology and Cognitive Science Center, National Taiwan University, Taipei, Taiwan, ⁹Center for Artificial Intelligence and Advanced Robotics, National Taiwan University, Taipei, Taiwan

Ming-Chih Kuo  
Graduate Institute of Medicine, Yuan Ze University|Department of Medical Imaging, National Taiwan University Cancer Center
Taoyuan, Taiwan|Taipei, Taiwan

Yao-Chia Shih  
Graduate Institute of Medicine, Yuan Ze University
Taoyuan, Taiwan

Ru-Jen Lin  
Department of Neurology, National Taiwan University Hospital Hsin-Chu Branch
Hsinchu City, Taiwan

Kai-Hsiang Chen  
Department of Neurology, National Taiwan University Hospital Hsin-Chu Branch
Hsinchu City, Taiwan

Yuh-Fen Wei  
Department of Medical Imaging, National Taiwan University Hospital Hsin-Chu Branch
Hsinchu City, Taiwan

Yen-Chung Hsiue  
Department of Medical Imaging, National Taiwan University Hospital Hsin-Chu Branch
Hsinchu City, Taiwan

Hui-Yu Yang  
Department of Neurology, National Taiwan University Hospital Hsin-Chu Branch
Hsinchu City, Taiwan

Xue-Zhe Lu  
Siemens Healthineers Limited
Taipei, Taipei

Joshua Oon Soo Goh  
Graduate Institute of Brain and Mind Sciences, College of Medicine, National Taiwan University|Department of Psychology, National Taiwan University|Neurobiology and Cognitive Science Center, National Taiwan University|Center for Artificial Intelligence and Advanced Robotics, National Taiwan University
Taipei, Taiwan|Taipei, Taiwan|Taipei, Taiwan|Taipei, Taiwan

Neuronal loss in the substantia nigra pars compacta (SNpc) and/or locus coeruleus (LC) is seen in Parkinson's disease (PD) (Braak, Del Tredici et al. 2003), resulting in a visible loss of neuromelanin (NM) detected by NM-MRI (Sasaki, Shibata et al. 2006). The contrast of NM-MRI is mainly caused by magnetization transfer (MT) and T1-shortening effects. But a conventional gradient-echo-MT sequence requires longer scan time (>8min). Two novel NM-MRI sequences with shorter scan time (<6min) have shown the comparable or better contrast-to-noise ratio (CNR) of NM-MRI signals in SNpc/LC than the conventional gradient-echo-MT one, namely black-blood delay alternating with nutation for tailored excitation-prepared T1-weighted variable flip angle turbo spin echo (DANTE T1-SPACE) (Oshima, Fushimi et al. 2021) and Sandwich-NM (Ji, Choi et al. 2022). Whilst the hyperintense signals in SNpc/LC and suppressed cerebrospinal fluid and intravascular signals shown in DANTE T1-SPACE images rely on MT effect due to short rectangular pulses in wideband, Sandwich-NM adopts even number of flow saturation pulses inferior and superior to the imaging slab to enhance incidental MT effect on NM CNR without flow artifacts. However, it is underdetermined which novel NM-MRI sequence within the above two has better performance on the diagnosis of PD. Thus, we simultaneously applied both sequences to each participant within a single MRI session, and aimed to compare their SNpc/LC CNR and PD diagnostic performance.

Twenty-four patients with PD (age=64.64±8.76 years; 14 males) and 24 age- and sex-matched healthy controls (HCs, age=67.46±7.72 years, 10 males) underwent all structural brain scans on a 3T-MRI scanner (MAGNETOM Skyra, Siemens Healthineers, Erlangen, Germany) with a 20‐channel head coil. The details of imaging parameters for both sequences are described in previous studies (Oshima, Fushimi et al. 2021, Ji, Choi et al. 2022). We used a MATLAB toolkit (SPM 12) to co-registered both DANTE T1-SPACE and Sandwich-NM images to a T1-weighted MPRAGE image for each subject. Two radiographers used 3D slicer software to manually draw regions of interest (ROIs) on the bilateral SNpc/LC on co-registered Sandwich-NM images. They also placed 3 circle reference ROIs on the bilateral cerebral crus (CC) and a non-SN brainstem tissue for each individual image. Both co-registered NM-MRI images thereby shared the same manually drawn ROIs that allowed us fairly calculate CNR from them. CNRSNpc in each side was calculated by the following equation: (SI_SNpc-SI_CC)/SI_CC, whereas CNR_LC was obtained by (SI_LC-SI_non-SN)/SI_non-SN (SI: averaged signal intensity within a ROI). Finally, a receiver operating characteristic curve (ROC) analysis was performed to differentiate patients from HCs in terms of either CNR_SNpc or CNR_LC in the left, right, and bilateral sides, to compare the diagnosis performance between both NM-MRI sequences.

Significant CNR reductions in the left, right, and bilateral SNpc in both DANTE T1-SPACE and Sandwich-NM images were found in PD compared to HC (p<0.001). However, CNR_LC in either of two sides disabled the differentiation between PD and HC groups. No significant differences in CNR measures between two raters. Results of ROC analysis showed that the CNR_SNpc obtained from both sides achieved great performance on discriminating PD and HC groups when using either DANTE T1-SPACE (area under the curve [AUC]=0.892) or Sandwich-NM (AUC=0.889). CNR_LC from either side failed to classify PD and HC groups (Fig. 2).

Our findings revealed that CNR_SNpc measures from DANTE T1-SPACE and Sandwich-NM images have comparable PD diagnostic performance, with nearly short scan time. Future works will use atlas-based ROI (Pauli, Nili et al. 2018) or deep-learning based segmentation (Le Berre, Kamagata et al. 2019) methods to improve CNR calculation, especially for potentially subjective biases due to tiny LC structure.

Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) ¹

Aging

Anatomical MRI

Neurophysiology of Imaging Signals ²

Brainstem

Movement Disorder

MRI

Neurological

Neuron

STRUCTURAL MRI

Other - Neuromelanin