Low-count whole-body PET with deep learning in a multicenter and externally validated study

Abstract

More widespread use of positron emission tomography (PET) imaging is limited by its high cost and radiation dose. Reductions in PET scan time or radiotracer dosage typically degrade diagnostic image quality (DIQ). Deep-learning-based reconstruction may improve DIQ, but such methods have not been clinically evaluated in a realistic multicenter, multivendor environment. In this study, we evaluated the performance and generalizability of a deep-learning-based image-quality enhancement algorithm applied to fourfold reduced-count whole-body PET in a realistic clinical oncologic imaging environment with multiple blinded readers, institutions, and scanner types. We demonstrate that the low-count-enhanced scans were noninferior to the standard scans in DIQ (p < 0.05) and overall diagnostic confidence (p < 0.001) independent of the underlying PET scanner used. Lesion detection for the low-count-enhanced scans had a high patient-level sensitivity of 0.94 (0.83–0.99) and specificity of 0.98 (0.95–0.99). Interscan kappa agreement of 0.85 was comparable to intrareader (0.88) and pairwise inter-reader agreements (maximum of 0.72). SUV quantification was comparable in the reference regions and lesions (lowest p-value=0.59) and had high correlation (lowest CCC = 0.94). Thus, we demonstrated that deep learning can be used to restore diagnostic image quality and maintain SUV accuracy for fourfold reduced-count PET scans, with interscan variations in lesion depiction, lower than intra- and interreader variations. This method generalized to an external validation set of clinical patients from multiple institutions and scanner types. Overall, this method may enable either dose or exam-duration reduction, increasing safety and lowering the cost of PET imaging.

Evaluation of a High-Sensitivity Organ-Targeted PET Camera

“…Despite the proven value of using radiotracers in a broad spectrum of diagnostic procedures across oncology, cardiology, and neurology, the standard radiotracer doses used in PET diagnostic procedures of 185 to 370 MBq [32] continues to define the limitations of PET imaging for use in undiagnosed patients (including screening procedures) and radiation-sensitive population [33]. In this regard, much attention is currently being paid in the literature to synthesizing high-quality PET images from input images acquired with a low dose of radiotracers through the use of deep learning and convolutional neural networks (CNNs) [16,34,35]. Despite the high potential of deep learning to denoise low-count PET images, a comprehensive approach is needed so that advances in software are complemented by improved PET-detector sensitivity to make low-dose PET imaging a clinical reality.…”

Deep learning-based image reconstruction and post-processing methods in positron emission tomography for low-dose imaging and resolution enhancement

“…Additional studies to evaluate the performance of deep learning-based methods across multiple sites will be crucial to demonstrate the performance of commercially viable deep learning-based methods. While those works which evaluated performance across sites [ 100 , 101 ] showed promising results, large-scale validation on a range of tracers and anatomical regions will be required to identify the shortcomings of deep learning-based methods in a clinical setting. Such studies will inform the development of commercially available deep learning-based software which should perform consistently across sites.…”

“…Results demonstrated improved lesion quantitation and detectability with radiation dose reductions of 50%. Chaudhari et al [ 100 ] evaluated the performance of a 2.5D Unet on whole body 18 F -FDG scans collected across three sites with a dose reduction factor of 4 × (Fig. 4 ).…”

“…4 Examples of deep learning-based low-dose to full-dose post-processing using PET only inputs and evaluated across multiple sites using different acquisition protocols and scanners. The neural network used in this case was trained using data sourced independently from the three evaluation sites to provide an unbiased evaluation (Image from [ 100 ]) …”

MRSaiFE: An AI-Based Approach Towards the Real-Time Prediction of Specific Absorption Rate

“…MR thermometry as an alternative approach suffers from a coarse temperature resolution [22]. The advent of artificial intelligence (AI) and machine learning in MRI [23][24][25][26][27][28][29][30][31][32][33][34][35][36] has opened up new avenues for the prediction of various imaging characteristics, among them the recent prediction of local SAR in prostate imaging [37][38][39][40], as well as the prediction of temperature rise in the brain for 33 different tissue types [41].…”