1
|
Li Y, Hu J, Sari H, Xue S, Ma R, Kandarpa S, Visvikis D, Rominger A, Liu H, Shi K. A deep neural network for parametric image reconstruction on a large axial field-of-view PET. Eur J Nucl Med Mol Imaging 2023; 50:701-714. [PMID: 36326869 DOI: 10.1007/s00259-022-06003-4] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 06/22/2022] [Accepted: 10/09/2022] [Indexed: 11/06/2022]
Abstract
PURPOSE The PET scanners with long axial field of view (AFOV) having ~ 20 times higher sensitivity than conventional scanners provide new opportunities for enhanced parametric imaging but suffer from the dramatically increased volume and complexity of dynamic data. This study reconstructed a high-quality direct Patlak Ki image from five-frame sinograms without input function by a deep learning framework based on DeepPET to explore the potential of artificial intelligence reducing the acquisition time and the dependence of input function in parametric imaging. METHODS This study was implemented on a large AFOV PET/CT scanner (Biograph Vision Quadra) and twenty patients were recruited with 18F-fluorodeoxyglucose (18F-FDG) dynamic scans. During training and testing of the proposed deep learning framework, the last five-frame (25 min, 40-65 min post-injection) sinograms were set as input and the reconstructed Patlak Ki images by a nested EM algorithm on the vendor were set as ground truth. To evaluate the image quality of predicted Ki images, mean square error (MSE), peak signal-to-noise ratio (PSNR), and structural similarity index measure (SSIM) were calculated. Meanwhile, a linear regression process was applied between predicted and true Ki means on avid malignant lesions and tumor volume of interests (VOIs). RESULTS In the testing phase, the proposed method achieved excellent MSE of less than 0.03%, high SSIM, and PSNR of ~ 0.98 and ~ 38 dB, respectively. Moreover, there was a high correlation (DeepPET: [Formula: see text]= 0.73, self-attention DeepPET: [Formula: see text]=0.82) between predicted Ki and traditionally reconstructed Patlak Ki means over eleven lesions. CONCLUSIONS The results show that the deep learning-based method produced high-quality parametric images from small frames of projection data without input function. It has much potential to address the dilemma of the long scan time and dependency on input function that still hamper the clinical translation of dynamic PET.
Collapse
Affiliation(s)
- Y Li
- College of Biomedical Engineering & Instrument Science, Zhejiang University, Hangzhou, People's Republic of China.,College of Optical Science and Engineering, Zhejiang University, Hangzhou, People's Republic of China
| | - J Hu
- Department of Nuclear Medicine, Inselpital, Bern University Hospital, University of Bern, Bern, Switzerland
| | - H Sari
- Advanced Clinical Imaging Technology, Siemens Healthcare AG, Lausanne, Switzerland
| | - S Xue
- Department of Nuclear Medicine, Inselpital, Bern University Hospital, University of Bern, Bern, Switzerland
| | - R Ma
- Department of Nuclear Medicine, Inselpital, Bern University Hospital, University of Bern, Bern, Switzerland.,Department of Engineering Physics, Tsinghua University, Beijing, China
| | - S Kandarpa
- LaTIM, INSERM, UMR 1101, University of Brest, Brest, France
| | - D Visvikis
- LaTIM, INSERM, UMR 1101, University of Brest, Brest, France
| | - A Rominger
- Department of Nuclear Medicine, Inselpital, Bern University Hospital, University of Bern, Bern, Switzerland
| | - H Liu
- College of Optical Science and Engineering, Zhejiang University, Hangzhou, People's Republic of China.
| | - K Shi
- Department of Nuclear Medicine, Inselpital, Bern University Hospital, University of Bern, Bern, Switzerland.,Computer Aided Medical Procedures and Augmented Reality, Institute of Informatics I16, Technical University of Munich, Munich, Germany
| |
Collapse
|
2
|
Santarelli MF, Genovesi D, Scipioni M, Positano V, Favilli B, Giorgetti A, Vergaro G, Landini L, Emdin M, Marzullo P. Cardiac amyloidosis characterization by kinetic model fitting on [18F]florbetaben PET images. J Nucl Cardiol 2022; 29:1919-1932. [PMID: 33864226 DOI: 10.1007/s12350-021-02608-8] [Citation(s) in RCA: 5] [Impact Index Per Article: 2.5] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 07/23/2020] [Accepted: 03/11/2021] [Indexed: 11/24/2022]
Abstract
OBJECTIVE To evaluate the feasibility of kinetic modeling-based approaches from [18F]-Flobetaben dynamic PET images as a non-invasive diagnostic method for cardiac amyloidosis (CA) and to identify the two AL- and ATTR-subtypes. METHODS AND RESULTS Twenty-one patients with diagnoses of CA (11 patients with AL-subtype and 10 patients with ATTR-subtype of CA) and 15 Control patients with no-CA conditions underwent PET/CT imaging after [18F]Florbetaben bolus injection. A two-tissue-compartment (2TC) kinetic model was fitted to time-activity curves (TAC) obtained from left ventricle wall and left atrium cavity ROIs to estimate kinetic micro- and macro-parameters. Combinations of kinetic parameters were evaluated with the purpose of distinguishing Control subjects and CA patients, and to correctly label the last ones as AL- or ATTR-subtype. Resulting sensitivity, specificity, and accuracy for Control subjects were: 0.87, 0.9, 0.89; as far as CA patients, the sensitivity, specificity, and accuracy were respectively 0.9, 1, and 0.97 for AL-CA patients and 0.9, 0.92, 0.97 for ATTR-CA patients. CONCLUSION Pharmacokinetic analysis based on a 2TC model allows cardiac amyloidosis characterization from dynamic [18F]Florbetaben PET images. Estimated model parameters allows to not only distinguish between Control subjects and patients, but also between AL- and ATTR-amyloid patients.
Collapse
Affiliation(s)
- M F Santarelli
- CNR Institute of Clinical Physiology, CNR Research Area - Via Moruzzi, 1, 56124, Pisa, Italy.
- Fondazione Toscana "G. Monasterio", Pisa, Italy.
| | - D Genovesi
- Fondazione Toscana "G. Monasterio", Pisa, Italy
| | - M Scipioni
- CNR Institute of Clinical Physiology, CNR Research Area - Via Moruzzi, 1, 56124, Pisa, Italy
- Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital and Harvard Medical School, Boston, MA, USA
| | - V Positano
- Fondazione Toscana "G. Monasterio", Pisa, Italy
| | - B Favilli
- Fondazione Toscana "G. Monasterio", Pisa, Italy
| | - A Giorgetti
- Fondazione Toscana "G. Monasterio", Pisa, Italy
| | - G Vergaro
- Scuola Universitaria Superiore 'S. Anna", Pisa, Italy
| | - L Landini
- Fondazione Toscana "G. Monasterio", Pisa, Italy
- Dipartimento di Ingegneria dell'Informazione: DII, Pisa University, Pisa, Italy
| | - M Emdin
- Fondazione Toscana "G. Monasterio", Pisa, Italy
- Scuola Universitaria Superiore 'S. Anna", Pisa, Italy
| | - P Marzullo
- Fondazione Toscana "G. Monasterio", Pisa, Italy
| |
Collapse
|
3
|
Regional Characterization of the Gottingen Minipig Brain by [18 F]FDG Dynamic Pet Modeling. J Med Biol Eng 2022. [DOI: 10.1007/s40846-022-00739-y] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 10/16/2022]
Abstract
Abstract
Purpose
To determine the best kinetic model to be applied on dynamic brain [18 F]FDG PET images by characterizing the regional brain glucose metabolism of normal Göttingen minipigs.
Methods
Nine Göttingen minipigs were scanned with a clinical PET/CT tomograph, starting from the injection of an intravenous bolus of [18 F]FDG, for about 25 min. Dynamic images were reconstructed and nine brain regions of interest (ROI), plus a vascular region, were defined and time-activity curves (TAC) were determined.
Three kinetic models were considered for fitting with experimental TACs: one-tissue compartment model 1TC, two-tissue irreversible compartment model 2TCi and two-tissue reversible model 2TC. Akaike Information Criterion was considered to evaluate the goodness of each model fitting. Regional and global kinetic parameter values were evaluated, in addition to the partition coefficient, net influx rate and retention index (RI).
Results
Both 2TCi and 2TC models turned out to be good choices for the next analysis. Parameter values were very similar between the different brain regions, with similar values to when the brain as a whole is considered (kinetic parameters mean values, from 2TCi model: K1 = 1.0 ml/g/min, k2 = 0.49 min− 1, k3 = 0.034 min− 1, K1/k2 = 2.14ml/g, Ki =0.069 ml/g/min; from 2TC model: K1 = 1.10 ml/g/min, k2 = 0.54 min− 1, k3 = 0.058 min− 1, k4 = 0.039 min− 1, K1/k2 = 2.18 ml/g, Ki = 0.10 ml/g/min; RI mean ± sd: 0.147 ± 0.037 min− 1), with the exception of the cerebellum (mean values from the 2TCi model: K1 = 0.52 ml/g/min, k2 = 0.56 min− 1, k3 = 0.025 min− 1, K1/k2 = 0.98ml/g, Ki=0.022 ml/g/min; from 2TC model: K1 = 0.54 ml/g/min, k2 = 0.61 min− 1, k3 = 0.044 min− 1, k4 = 0.038 min− 1, K1/k2 = 0.95ml/g, Ki=0.032 ml/g/min; RI mean ± sd: 0.071 ± 0.018 min− 1).
Conclusion
The two-tissue model is able to describe the regional brain metabolism in Göttingen minipigs. Compared to the 2TCi model, in the 2TC model the k4 micro-parameter was also evaluated. This led to adjustments of the other microparameters, especially k3 and consequently the net influx rate Ki. For healthy minipigs, the glucose metabolism was similar in all of the brain regions analyzed, with the exception of the cerebellum, where the FDG uptake was lower.
Collapse
|
4
|
Santarelli MF, Scipioni M, Genovesi D, Giorgetti A, Marzullo P, Landini L. Imaging Techniques as an Aid in the Early Detection of Cardiac Amyloidosis. Curr Pharm Des 2021; 27:1878-1889. [PMID: 32787756 DOI: 10.2174/1381612826666200813133557] [Citation(s) in RCA: 5] [Impact Index Per Article: 1.7] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 03/06/2020] [Accepted: 06/23/2020] [Indexed: 11/22/2022]
Abstract
The idea that performing a proper succession of imaging tests and techniques allows an accurate and early diagnosis of cardiac amyloidosis, avoiding the need to perform the myocardial biopsy, is becoming increasingly popular. Furthermore, being imaging techniques non-invasive, it is possible to perform the follow-up of the pathology through repeated image acquisitions. In the present review, the various innovative imaging methodologies are presented, and it is discussed how they have been applied for early diagnosis of cardiac amyloidosis (CA), also to distinguish the two most frequent subtypes in CA: immunoglobulin light chain amyloidosis (AL) and transthyretin amyloidosis (ATTR); this allows to perform the therapy in a targeted and rapid manner.
Collapse
Affiliation(s)
| | - M Scipioni
- CNR Institute of Clinical Physiology, Pisa, Italy
| | - D Genovesi
- Fondazione Toscana "G. Monasterio", Pisa, Italy
| | - A Giorgetti
- Fondazione Toscana "G. Monasterio", Pisa, Italy
| | - P Marzullo
- Fondazione Toscana "G. Monasterio", Pisa, Italy
| | - L Landini
- Fondazione Toscana "G. Monasterio", Pisa, Italy
| |
Collapse
|
5
|
Scipioni M, Pedemonte S, Santarelli MF, Landini L. Probabilistic Graphical Models for Dynamic PET: A Novel Approach to Direct Parametric Map Estimation and Image Reconstruction. IEEE TRANSACTIONS ON MEDICAL IMAGING 2020; 39:152-160. [PMID: 31199257 DOI: 10.1109/tmi.2019.2922448] [Citation(s) in RCA: 3] [Impact Index Per Article: 0.8] [Reference Citation Analysis] [Abstract] [MESH Headings] [Track Full Text] [Subscribe] [Scholar Register] [Indexed: 06/09/2023]
Abstract
In the context of dynamic emission tomography, the conventional processing pipeline consists of independent image reconstruction of single-time frames, followed by the application of a suitable kinetic model to time-activity curves (TACs) at the voxel or region-of-interest level. Direct 4D positron emission tomography (PET) reconstruction, by contrast, seeks to move beyond this scheme and incorporate information from multiple time frames within the reconstruction task. Established direct methods are based on a deterministic description of voxelwise TACs, captured by the chosen kinetic model, considering the photon counting process the only source of uncertainty. In this paper, we introduce a new probabilistic modeling strategy based on the key assumption that activity time course would be subject to uncertainty even if the parameters of the underlying dynamic process are known. This leads to a hierarchical model that we formulate using the formalism of probabilistic graphical modeling. The inference is addressed using a new iterative algorithm, in which kinetic modeling results are treated as prior expectation of activity time course, rather than as a deterministic match, making it possible to control the trade-off between a data-driven and a model-driven reconstruction. The proposed method is flexible to an arbitrary choice of (linear and nonlinear) kinetic models, it enables the inclusion of arbitrary (sub)differentiable priors for parametric maps, and it is simple to implement. Computer simulations and an application to a real-patient scan show how the proposed method is able to generalize over conventional indirect and direct approaches, providing a bridge between them by properly tuning the impact of the kinetic modeling step on image reconstruction.
Collapse
|
6
|
Scipioni M. Direct 4D PET reconstruction with discrete tissue types. ANNUAL INTERNATIONAL CONFERENCE OF THE IEEE ENGINEERING IN MEDICINE AND BIOLOGY SOCIETY. IEEE ENGINEERING IN MEDICINE AND BIOLOGY SOCIETY. ANNUAL INTERNATIONAL CONFERENCE 2019; 2019:4840-4843. [PMID: 31946945 DOI: 10.1109/embc.2019.8856326] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [MESH Headings] [Track Full Text] [Subscribe] [Scholar Register] [Indexed: 06/10/2023]
Abstract
Dynamic positron emission tomography (dPET) is known for its ability to extract spatiotemporal information of a radio tracer in living tissue. In this paper, a novel direct reconstruction framework is presented, which include concurrent clustering as a potential aid in addressing high levels of noise typical of voxel-wise kinetic modeling. Core assumption is that the imaged volume is formed by a finite number of different functional regions, and that voxel-wise time courses are determined by the functional cluster they belong to. Probabilistic Graphical Modeling (PGM) theory is used to describe the problem, and to derive the inference strategy. The proposed iterative estimation scheme provides concurrent estimate of kinetic parameter maps, activity images, and segmented clusters. Simulation studies and exploratory application to real data are performed to validate the proposal.
Collapse
|