Deep learning-based time-of-flight (ToF) image enhancement of non-ToF PET scans

被引:35
作者
Mehranian, Abolfazl [1 ]
Wollenweber, Scott D. [2 ]
Walker, Matthew D. [3 ]
Bradley, Kevin M. [4 ]
Fielding, Patrick A. [5 ]
Huellner, Martin [6 ]
Kotasidis, Fotis [7 ]
Su, Kuan-Hao [2 ]
Johnsen, Robert [2 ]
Jansen, Floris P. [2 ]
McGowan, Daniel R. [3 ,8 ]
机构
[1] Univ Oxford, Big Data Inst, GE Healthcare, Oxford, England
[2] GE Healthcare, Waukesha, WI USA
[3] Oxford Univ Hosp NHS FT, Dept Med Phys & Clin Engn, Oxford, England
[4] Univ Hosp Wales, Wales Res & Diagnost PET Imaging Ctr, Cardiff, Wales
[5] Univ Hosp Wales, Dept Radiol, Cardiff, Wales
[6] Zurich Univ Hosp, Zurich, Switzerland
[7] GE Healthcare, Zurich, Switzerland
[8] Univ Oxford, Dept Oncol, Oxford, England
基金
“创新英国”项目;
关键词
Deep neural networks; Time of flight; PET; Image quality; RECONSTRUCTION;
D O I
10.1007/s00259-022-05824-7
中图分类号
R8 [特种医学]; R445 [影像诊断学];
学科分类号
1002 ; 100207 ; 1009 ;
摘要
Purpose To improve the quantitative accuracy and diagnostic confidence of PET images reconstructed without time-of-flight (ToF) using deep learning models trained for ToF image enhancement (DL-ToF). Methods A total of 273 [F-18]-FDG PET scans were used, including data from 6 centres equipped with GE Discovery MI ToF scanners. PET data were reconstructed using the block-sequential-regularised-expectation-maximisation (BSREM) algorithm with and without ToF. The images were then split into training (n = 208), validation (n = 15), and testing (n = 50) sets. Three DL-ToF models were trained to transform non-ToF BSREM images to their target ToF images with different levels of DL-ToF strength (low, medium, high). The models were objectively evaluated using the testing set based on standardised uptake value (SUV) in 139 identified lesions, and in normal regions of liver and lungs. Three radiologists subjectively rated the models using testing sets based on lesion detectability, diagnostic confidence, and image noise/quality. Results The non-ToF, DL-ToF low, medium, and high methods resulted in - 28 +/- 18, - 28 +/- 19, - 8 +/- 22, and 1.7 +/- 24% differences (mean; SD) in the SUVmax for the lesions in testing set, compared to ToF-BSREM image. In background lung VOIs, the SUVmean differences were 7 +/- 15, 0.6 +/- 12, 1 +/- 13, and 1 +/- 11% respectively. In normal liver, SUVmean differences were 4 +/- 5, 0.7 +/- 4, 0.8 +/- 4, and 0.1 +/- 4%. Visual inspection showed that our DL-ToF improved feature sharpness and convergence towards ToF reconstruction. Blinded clinical readings of testing sets for diagnostic confidence (scale 0-5) showed that non-ToF, DL-ToF low, medium, and high, and ToF images scored 3.0, 3.0, 4.1, 3.8, and 3.5 respectively. For this set of images, DL-ToF medium therefore scored highest for diagnostic confidence. Conclusion Deep learning-based image enhancement models may provide converged ToF-equivalent image quality without ToF reconstruction. In clinical scoring DL-ToF-enhanced non-ToF images (medium and high) on average scored as high as, or higher than, ToF images. The model is generalisable and hence, could be applied to non-ToF images from BGO-based PET/CT scanners.
引用
收藏
页码:3740 / 3749
页数:10
相关论文
共 22 条
[1]   The promise of artificial intelligence and deep learning in PET and SPECT imaging [J].
Arabi, Hossein ;
AkhavanAllaf, Azadeh ;
Sanaat, Amirhossein ;
Shiri, Isaac ;
Zaidi, Habib .
PHYSICA MEDICA-EUROPEAN JOURNAL OF MEDICAL PHYSICS, 2021, 83 :122-137
[2]   Using convolutional neural networks to estimate time-of-flight from PET detector waveforms [J].
Berg, Eric ;
Cherry, Simon R. .
PHYSICS IN MEDICINE AND BIOLOGY, 2018, 63 (02)
[3]  
Cheng L, 2017, INT C FULL 3 DIM IM
[4]   Effect of a Bayesian Penalized Likelihood PET Reconstruction Compared With Ordered Subset Expectation Maximization on Clinical Image Quality Over a Wide Range of Patient Weights [J].
Chilcott, Anna K. ;
Bradley, Kevin M. ;
McGowan, Daniel R. .
AMERICAN JOURNAL OF ROENTGENOLOGY, 2018, 210 (01) :153-157
[5]   Focus on time-of-flight PET: the benefits of improved time resolution [J].
Conti, Maurizio .
EUROPEAN JOURNAL OF NUCLEAR MEDICINE AND MOLECULAR IMAGING, 2011, 38 (06) :1147-1157
[6]   Why is TOF PET reconstruction a more robust method in the presence of inconsistent data? [J].
Conti, Maurizio .
PHYSICS IN MEDICINE AND BIOLOGY, 2011, 56 (01) :155-168
[7]   Improvement in Lesion Detection with Whole-Body Oncologic Time-of-Flight PET [J].
El Fakhri, Georges ;
Surti, Suleman ;
Trott, Cathryn M. ;
Scheuermann, Joshua ;
Karp, Joel S. .
JOURNAL OF NUCLEAR MEDICINE, 2011, 52 (03) :347-353
[8]   Machine Learning in PET: From Photon Detection to Quantitative Image Reconstruction [J].
Gong, Kuang ;
Berg, Eric ;
Cherry, Simon R. ;
Qi, Jinyi .
PROCEEDINGS OF THE IEEE, 2020, 108 (01) :51-68
[9]   Benefit of time-of-flight in PET: Experimental and clinical results [J].
Karp, Joel S. ;
Surti, Sulernan ;
Daube-Witherspoon, Margaret E. ;
Muehllehner, Gerd .
JOURNAL OF NUCLEAR MEDICINE, 2008, 49 (03) :462-470
[10]  
Kingma D. P., 2015, ACS SYM SER