Dataset Information

Estimation of pharmacokinetic parameters from DCE-MRI by extracting long and short time-dependent features using an LSTM network.

ABSTRACT:

Purpose

T₁ -weighted dynamic contrast-enhanced Magnetic Resonance Imaging (DCE-MRI) is typically quantified by least squares (LS) fitting to a pharmacokinetic (PK) model to yield parameters of microvasculature and perfusion in normal and disease tissues. Such fitting is both time-consuming as well as subject to inaccuracy and instability in parameter estimates. Here, we propose a novel neural network approach to estimate the PK parameters by extracting long and short time-dependent features in DCE-MRI.

Methods

A Long Short-Term Memory (LSTM) network, widely used for processing sequence data, was employed to map DCE-MRI time-series accompanied with an arterial input function to parameters of the extended Tofts model. Head and neck DCE-MRI from 103 patients were used for training and testing the LSTM model. Arterial input functions (AIFs) from 78 patients were used to generate synthetic DCE-MRI time-series for training, during which data augmentation was used to overcome the limited size of in vivo data. The model was tested on independent synthesized DCE data using AIFs from 25 patients. The LSTM performance was optimized for the numbers of layers and hidden state features. The performance of the LSTM was tested for different temporal resolution, total acquisition time, and contrast-to-noise ratio (CNR), and compared to the conventional LS fitting and a CNN-based method.

Results

Compared to LS fitting, the LSTM model had comparable accuracy in PK parameter estimations from fully temporal-sampled DCE-MRI data (~3 s per frame), but much better accuracy for the data with temporally subsampling (4s or greater per frame), total acquisition time truncation by 48%-16%, or low CNR (5 and 10). The LSTM reduced normalized root mean squared error by 40.4%, 46.9%, and 53.0% for sampling intervals of 4s, 5s, and 6s, respectively, compared to LS fitting. Compared to the CNN model, the LSTM model reduced the error in the parameter estimates up to 55.2%. Also, the LSTM improved the inference time by ~ 14 times on CPU compared to LS fitting.

Conclusion

Our study suggests that the LSTM model could achieve improved robustness and computation speed for PK parameter estimation compared to LS fitting and the CNN based network, particularly for suboptimal data.

SUBMITTER: Zou J

PROVIDER: S-EPMC7429301 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:PurposeTumor hypoxia is a major cause of radiation resistance, often present in various solid tumors. Dynamic [18 F]-fluoromisonidazole (FMISO) PET imaging is able to reliably assess tumor hypoxia. Comprehensive characterization of tumor microenvironment through FMISO-PET and dynamic contrast enhanced (DCE) MR multimodality imaging might be a valuable alternative to the dynamic FMISO-PET acquisition. The aim of this work was to explore the correlation between the FMISO-PET and DCE-MRI kinetic parameters.MethodsThis study was done on head and neck cancer patients (N = 6), who were imaged dynamically with FMISO-PET and DCE-MRI on the same day. Images were registered and analyzed for kinetics on a voxel basis. FMISO-PET images were analyzed with the two-tissue compartment three rate-constant model. Additionally, tumor-to-muscle ratio (TMR) maps were evaluated. DCE-MRI was analyzed with the extended Tofts model. Voxel-wise Pearson's coefficients were calculated for each patient to assess pairwise parameter correlations.ResultsMedian correlations between FMISO uptake parameters and DCE-MRI kinetic parameters varied across the parameter pairs in the range from -0.05 to 0.71. The highest median correlation of r = 0.71 was observed for the pair Vb -vp , while the K1 -Ktrans median correlation was r = 0.45. Median correlation coefficients for the K1 -vp and the Ki -Ktrans pairs were r = 0.42 and r = 0.32, respectively. Correlations between FMISO uptake rate parameter Ki and DCE-MRI kinetic parameters varied substantially across the patients, whereas correlations between the FMISO and DCE-MRI vascular parameters were consistently high. Median TMR-K1 and TMR-Ktrans correlations were r = 0.52 and r = 0.46, respectively, but varied substantially across the patients.ConclusionsBased on this clinical evidence, we can conclude that the vascular fraction parameters obtained through DCE-MRI kinetic analysis or FMISO kinetic analysis measure the same biological property, while other kinetic parameters are unrelated. These results might be useful in the design of future clinical trials involving FMISO-PET/DCE-MR multimodality imaging for the assessment of tumor microenvironment.

Project description:The aim of the study was to investigate dynamic contrast enhanced MRI (DCE-MRI) as a potential marker to assess the therapeutic responses of fecal microbiota transplantation (FMT) in patients with Crohn's disease (CD) and to determine the parameter or combination of parameters most strongly associated with changes in clinical indicators after treatment.In 22 CD patients, DCE-MRI was performed with a 3.0T scanner. Parameters of DCE-MRI (vascular transfer constant [K] and blood volume [BV]) in the terminal ileum were compared between before and day 90 after FMT treatment. The differences of clinical indicators (C-reactive protein [CRP], Harvey-Bradshaw index [HBI]) and DCE-MRI parameters (K, BV) between pre- and post-treatment was calculated by Student's 2-tailed, paired t-test. The correlations between percent change of clinical indicators (ΔCRP, ΔHBI) with DCE-MRI parameters (ΔK, ΔBV) were analyzed by Pearson's correlation coefficients. A logistic regression model was used to identify the changes of DCE-MRI parameters related to the treatment outcomes. Receiver operating characteristic curves (ROCs) were generated to assess which DCE-MRI parameter showed the best accuracy for evaluation of therapeutic response.After treatment, mean values of clinical indicators decreased significantly (CRP: 62.68 ± 31.86 vs 43.55 ± 29.63 mg/L, P = .008; HBI: 7.18 ± 2.10 vs 5.73 ± 2.33, P = 0.012). Both DCE-MRI parameters showed prominent differences before and after treatment: K (1.86 ± 0.87 vs 1.39 ± 0.83 min, P = .017), BV (61.02 ± 28.49 vs 41.96 ± 22.75 mL/100 g, P = .005). There were significant correlations between ΔCRP or ΔHBI and percent change of CDE-MRI parameters (ΔK to ΔCRP: 0.659; ΔK to ΔHBI: 0.496; ΔBV to ΔCRP: 0.442; ΔBV to ΔHBI: 0.476). Compared to ΔK and ΔBV individually, the combination of both parameters performed best in assessment of therapeutic response with an area under the ROCs (AUC) of 0.948.K and BV parameters derived from DCE-MRI have the potential to assess for therapeutic response after FMT treatment for CD. The combination of K and BV measurements improved the predictive capability compared to the individual parameters.