Adaptive template generation for amyloid PET using a deep learning approach

Abstract

Accurate spatial normalization (SN) of amyloid positron emission tomography (PET) images for Alzheimer's disease assessment without coregistered anatomical magnetic resonance imaging (MRI) of the same individual is technically challenging. In this study, we applied deep neural networks to generate individually adaptive PET templates for robust and accurate SN of amyloid PET without using matched 3D MR images. Using 681 pairs of simultaneously acquired C-PIB PET and T1-weighted 3D MRI scans of AD, MCI, and cognitively normal subjects, we trained and tested two deep neural networks [convolutional auto-encoder (CAE) and generative adversarial network (GAN)] that produce adaptive best PET templates. More specifically, the networks were trained using 685,100 pieces of augmented data generated by rotating 527 randomly selected datasets and validated using 154 datasets. The input to the supervised neural networks was the 3D PET volume in native space and the label was the spatially normalized 3D PET image using the transformation parameters obtained from MRI-based SN. The proposed deep learning approach significantly enhanced the quantitative accuracy of MRI-less amyloid PET assessment by reducing the SN error observed when an average amyloid PET template is used. Given an input image, the trained deep neural networks rapidly provide individually adaptive 3D PET templates without any discontinuity between the slices (in 0.02 s). As the proposed method does not require 3D MRI for the SN of PET images, it has great potential for use in routine analysis of amyloid PET images in clinical practice and research.

A Review of Deep-Learning-Based Approaches for Attenuation Correction in Positron Emission Tomography

“…In most areas of nuclear medicine, DL-based image processing and analysis techniques have received significant research attention [184,195,196]; namely, the DL-based image reconstruction and denoising for radiation dose reduction [197][198][199][200], automatic segmentation of various organs and structures for quantitative image analyses [201,202], image spatial normalization [203,204], voxel-based internal dosimetry [205], and the image-to-image transition for PET AC. Moreover, the DL-based lesion detection and image interpretation received significant research attention [206][207][208][209].…”

Development and evaluation of a T1 standard brain template for Alzheimer disease

“…MRI scans for Cohorts 1 to 4 were acquired for each participant using a 3-T MR system (Achieva, Philips Medical Systems, Best, The Netherlands) equipped with an eight-channel sensitivity encoding head coil. A sagittal structural 3D T1-weighted (T1W) image was acquired using a turbo field echo sequence that is similar to the magnetization-prepared rapid acquisition of gradient echo (MPRAGE) sequence with the following parameters: repetition time (TR) =8.1 ms, echo time (TE) =3.7 ms, flip angle (FA) =8°, field-of-view (FOV) =236×236 mm 2 , acquisition voxel size =1×1×1 mm 3 , and reconstruction voxel size =1×1×1 mm 3 . In addition, T2-weighted turbo-spin-echo and FLAIR images were acquired to examine any brain malformations.…”

Improving PET Imaging Acquisition and Analysis With Machine Learning: A Narrative Review With Focus on Alzheimer's Disease and Oncology

“…Kang et al trained 2 CNNs: convolutional autoencoder and generative adversarial network, using a patient dataset (n ¼ 681) of simultaneously acquired [ 11 C]PiB PET/MR scans of individuals with AD, mild cognitive impairment (MCI), and healthy controls (HCs). 31 The training regimen produced a spatially normalized PET image, given the transformation parameters obtained by the SN generated from the accompanying MR data. As a result, given an inputted PET image in native space, the neural network was able to generate an individually adaptive amyloid PET template to achieve accurate SN without the use of MR data.…”