Summary: A new deep learning algorithm utilizes neuroimaging data to differentiate between Parkinson’s disease and other parkinsonian syndromes such as PSP and multiple systems atrophy.
A new deep learning method has been created to aid in the diagnosis of parkinsonian diseases, according to research published ahead of print by The Journal of Nuclear Medicine.
Using a 3D deep convolutional neural network to extract deep metabolic imaging indices from 18F-FDG PET scans, scientists can effectively differentiate between Parkinson’s disease and other parkinsonian syndromes, such as multiple system atrophy and progressive supranuclear palsy.
Parkinson’s disease is one of the most common neurodegenerative disorders. 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—particularly in the early stages—as its symptoms overlap considerably with those of other atypical parkinsonian syndromes.
“Studies show that 20 to 30 percent of patients with initial diagnoses of Parkinson’s disease were subsequently demonstrated to have multiple system atrophy or progressive supranuclear palsy after pathological examination,” said Ping Wu, MD, PhD, neuroradiologist at PET Center, Huashan Hospital, Fudan University in Shanghai, China.
“Therefore, the development of accurate indices to differentiate between parkinsonian diseases is of great importance, specifically with regard to determining treatment strategies.”
To achieve this objective, researchers built a 3D deep convolutional neural network, known as the Parkinsonism Differential Diagnosis Network (PDD-Net), to automatically identify imaging-related indices that could support the differential diagnosis of parkinsonian diseases. This deep learning method was used to examine parkinsonian PET imaging from two groups: more than 2,100 patients from China and 90 patients from Germany.
“It’s important to note the steps that were taken to improve the trustworthiness of the study,” said Wu. “We utilized the largest benchmark dataset of parkinsonian patients with FDG PET from Huashan Parkinsonian PET Imaging database in Shanghai, China, and conducted extensive testing on longitudinal data. In addition, we studied the German cohort to include external data representing different ethnicities and examination protocols.”
The 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 progressive supranuclear palsy.
“This work confirms that the emerging artificial intelligence can extract in-depth information from molecular imaging to enhance the differentiation of complex physiology,” Wu said. “Deep learning technology may help physicians maximize the utility of nuclear medicine imaging in the future.”
About this deep learning and Parkinson’s disease research news
Author: Rebecca Maxey Source: SNMMI Contact: Rebecca Maxey – SNMMI Image: The image is credited to Ping Wu and Yu Zhao, et al
Differential diagnosis of parkinsonism based on deep metabolic imaging indices
The clinical presentations of early idiopathic Parkinson’s disease (PD) substantially overlap with those of atypical parkinsonian syndromes like multiple system atrophy (MSA) and progressive supranuclear palsy (PSP). This study aimed to develop metabolic imaging indices based on deep learning to support the differential diagnosis of these conditions.
Methods: A benchmark Huashan parkinsonian PET imaging (HPPI, China) database including 1275 parkinsonian patients and 863 non-parkinsonian subjects with 18F-FDG PET images was established to support artificial intelligence development. A 3D deep convolutional neural network was developed to extract deep metabolic imaging (DMI) indices, which was blindly evaluated in an independent cohort with longitudinal follow-up from the HPPI, and an external German cohort of 90 parkinsonian patients with different imaging acquisition protocols.
Results: The proposed DMI indices had less ambiguity space in the differential diagnosis. They achieved sensitivities of 98.1%, 88.5%, and 84.5%, and specificities of 90.0%, 99.2%, and 97.8% for the diagnosis of PD, MSA, and PSP in the blind test cohort. In the German cohort, They resulted in sensitivities of 94.1%, 82.4%, 82.1%, and specificities of 84.0%, 99.9%, 94.1% respectively. Employing the PET scans independently achieved comparable performance to the integration of demographic and clinical information into the DMI indices.
Conclusion: The DMI indices developed on the HPPI database show potential to provide an early and accurate differential diagnosis for parkinsonism and is robust when dealing with discrepancies between populations and imaging acquisitions.