New Deep Learning Method Provides Early and Accurate Differential Diagnosis of Parkinsonian Diseases

Reston, VA-A new deep learning method has been created to aid in the diagnosis of Parkinsonian diseases, according to a study published ahead of print by The Journal of Nuclear Medicine. Using a 3D deep convolution neural network to extract deep metabolic imaging cues from 18Using F-FDG PET, scientists can effectively differentiate Parkinson’s disease from other parkinsonian syndromes, such as multiple system atrophy and progressive supranuclear palsy.

Parkinson’s disease is one of the most common neurodegenerative diseases. According to the Parkinson’s Foundation, more than 10 million people worldwide live with the disease. Accurate diagnosis of Parkinson’s disease is often a challenge, especially in the early stages, because its symptoms overlap considerably with those of other atypical parkinsonian syndromes.

“Studies show that 20-30% of patients initially diagnosed with Parkinson’s disease later presented with multiple system atrophy or progressive supranuclear palsy after pathological examination,” said Ping Wu, MD, PhD, neuroradiologist at PET Center, Huashan Hospital, Fudan. University of Shanghai, China. “Therefore, the development of accurate indices to differentiate Parkinsonian diseases is of great importance, especially with regard to determining treatment strategies.”

To achieve this goal, the researchers built a 3D deep convolution neural network, known as the Parkinsonism Differential Diagnosis Network (PDD-Net), to automatically identify imaging-related clues that could support the differential diagnosis of parkinsonian diseases. This deep learning method was used to examine the Parkinsonian PET imaging of two groups: more than 2,100 Chinese patients and 90 German patients.

“It is important to note the steps that were taken to improve the reliability of the study,” Wu said. Huashan Parkinsonian PET Imaging data center in Shanghai, China, and performed extensive tests on longitudinal data. Additionally, we studied the German cohort to include external data representing different ethnicities and examination protocols.

Deep metabolic imaging indices extracted from PDD-Net provided an early and accurate method for the differential diagnosis of parkinsonian syndromes, with high rates of sensitivity and specificity for Parkinson’s disease, multiple system atrophy and supranuclear palsy progressive.

“This work confirms that emerging artificial intelligence can extract detailed information from molecular imaging to improve the differentiation of complex physiology,” Wu said. “Deep learning technology can help physicians maximize the usefulness of imaging in nuclear medicine in the future.”

Supplemental Figure 4: Visualization of mean salience maps of patients with idiopathic Parkinson’s disease (IPD), multiple system atrophy (MSA), and progressive supranuclear palsy (PSP) in the training cohort showing characteristic regions contributing deep metabolic imaging (DMI) indices. The color corresponds to the importance score indicating the contribution of a region for the deep metabolic imaging (DMI) indices generated. The color directions (yellow and red versus cyan and blue) represent different influences on the DMI indices (an increased absorbance value contributes to the increase or decrease in the likelihood of IPD, MSA, or PSP in the DMI indices). The arrows pointed to the most prominent brain regions, including 1: cerebellum, 2: midbrain, 3: putamen, 4: thalamus.

The authors of “Differential Diagnosis of Parkinsonism Based on Deep Metabolic Imaging Clues” include Ping Wu, Yihui Guan, and Chuantao Zuo, PET Center, Huashan Hospital, Fudan University, Shanghai, China, and National Research Center for Aging and Medicine & National Center for Neurological Disorders, Huashan Hospital, Fudan University, Shanghai, China; Yu Zhao, Department of Nuclear Medicine, University of Bern, Bern, Switzerland, AI Lab, Tencent, Shenzhen, China, and Department of Computer Science, Technische Universität, München, Munich, Germany; Jianjun Wu, Department of Neurology, Huashan Hospital, Fudan University, Shanghai, China; Matthias Brendel, Department of Nuclear Medicine, University of Munich, Munich, Germany; Jiaying, Lu, PET Center, Huashan Hospital, Fudan University, Shanghai, China, National Research Center for Aging and Medicine & National Center for Neurological Disorders, Huashan Hospital, Fudan University, Shanghai, China, and Department of Nuclear Medicine, University of Bern , Bern, Switzerland; Jingjie Ge, Ling Li, and Qian Xu, PET Center, Huashan Hospital, Fudan University, Shanghai, China; Alexander Bernhardt, Sabrina Katzdobler and Johannes Levin, Department of Neurology, University of Munich, Munich, Germany; Ian Alberts, Jimin Hong and Axel Rominger, Department of Nuclear Medicine, University of Bern, Bern, Switzerland; Igor Yakushev and Wolfgang Weber, Department of Nuclear Medicine, Technische Universität, München, Munich, Germany; Yimin Sun, Fengtao Liu and Jian Wang, National Research Center for Aging and Medicine and National Center for Neurological Disorders, Huashan Hospital, Fudan University, Shanghai, China, and Department of Neurology, Huashan Hospital, Fudan University, Shanghai, China ; Günter U. Höglinger, Department of Neurology, Hanover Medical School, Hanover, Germany; Claudio Bassetti, Department of Neurology, University of Bern, Bern, Switzerland; Wolfgang H. Oertel, Department of Neurology, University of Marburg, Marburg, Germany; and Kuangyu Shi, Department of Nuclear Medicine, University of Bern, Bern, Switzerland and Department of Computer Science, Technische Universität München, Munich, Germany.

Visit the JNM website for the latest research and follow our new Twitter and @JournalofNucMed Facebook pages or follow us on LinkedIn.

/Public release. This material from the original organization/authors may be ad hoc in nature, edited for clarity, style and length. The views and opinions expressed are those of the author or authors. See in full here.

Back To Top