Dataset Information

Discriminating pseudoprogression and true progression in diffuse infiltrating glioma using multi-parametric MRI data through deep learning.

ABSTRACT: Differentiating pseudoprogression from true tumor progression has become a significant challenge in follow-up of diffuse infiltrating gliomas, particularly high grade, which leads to a potential treatment delay for patients with early glioma recurrence. In this study, we proposed to use a multiparametric MRI data as a sequence input for the convolutional neural network with the recurrent neural network based deep learning structure to discriminate between pseudoprogression and true tumor progression. In this study, 43 biopsy-proven patient data identified as diffuse infiltrating glioma patients whose disease progressed/recurred were used. The dataset consists of five original MRI sequences; pre-contrast T1-weighted, post-contrast T1-weighted, T2-weighted, FLAIR, and ADC images as well as two engineered sequences; T1post-T1pre and T2-FLAIR. Next, we used three CNN-LSTM models with a different set of sequences as input sequences to pass through CNN-LSTM layers. We performed threefold cross-validation in the training dataset and generated the boxplot, accuracy, and ROC curve, AUC from each trained model with the test dataset to evaluate models. The mean accuracy for VGG16 models ranged from 0.44 to 0.60 and the mean AUC ranged from 0.47 to 0.59. For CNN-LSTM model, the mean accuracy ranged from 0.62 to 0.75 and the mean AUC ranged from 0.64 to 0.81. The performance of the proposed CNN-LSTM with multiparametric sequence data was found to outperform the popular convolutional CNN with a single MRI sequence. In conclusion, incorporating all available MRI sequences into a sequence input for a CNN-LSTM model improved diagnostic performance for discriminating between pseudoprogression and true tumor progression.

SUBMITTER: Lee J

PROVIDER: S-EPMC7683728 | biostudies-literature | 2020 Nov

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Discriminating pseudoprogression and true progression in diffuse infiltrating glioma using multi-parametric MRI data through deep learning.

Lee Joonsang J Wang Nicholas N Turk Sevcan S Mohammed Shariq S Lobo Remy R Kim John J Liao Eric E Camelo-Piragua Sandra S Kim Michelle M Junck Larry L Bapuraj Jayapalli J Srinivasan Ashok A Rao Arvind A

Scientific reports 20201123 1

Differentiating pseudoprogression from true tumor progression has become a significant challenge in follow-up of diffuse infiltrating gliomas, particularly high grade, which leads to a potential treatment delay for patients with early glioma recurrence. In this study, we proposed to use a multiparametric MRI data as a sequence input for the convolutional neural network with the recurrent neural network based deep learning structure to discriminate between pseudoprogression and true tumor progres ...[more]

PMID: 33230285

Similar Datasets

Project description:Background and purposeImmune response to cancer therapy may result in pseudoprogression, which can only be identified retrospectively and may disrupt an effective therapy. This study assesses whether serial parametric response mapping (a voxel-by-voxel method of image analysis also known as functional diffusion mapping) analysis of ADC measurements following peptide-based vaccination may help prospectively distinguish progression from pseudoprogression in pediatric patients with diffuse intrinsic pontine gliomas.Materials and methodsFrom 2009 to 2012, 21 children, 4-18 years of age, with diffuse intrinsic pontine gliomas were enrolled in a serial peptide-based vaccination protocol following radiation therapy. DWI was acquired before immunotherapy and at 6-week intervals during vaccine treatment. Pseudoprogression was identified retrospectively on the basis of clinical and radiographic findings, excluding DWI. Parametric response mapping was used to analyze 96 scans, comparing ADC measures at multiple time points (from the first vaccine to up to 12 weeks after the vaccine was halted) with prevaccine baseline values. Log-transformed fractional increased ADC, fractional decreased ADC, and parametric response mapping ratio (fractional increased ADC/fractional decreased ADC) were compared between patients with and without pseudoprogression, by using generalized estimating equations with inverse weighting by cluster size.ResultsMedian survival was 13.1 months from diagnosis (range, 6.4-24.9 months). Four of 21 children (19%) were assessed as experiencing pseudoprogression. Patients with pseudoprogression had higher fitted average log-transformed parametric response mapping ratios (P = .01) and fractional decreased ADCs (P = .0004), compared with patients without pseudoprogression.ConclusionsSerial parametric response mapping of ADC, performed at multiple time points of therapy, may distinguish pseudoprogression from true progression in patients with diffuse intrinsic pontine gliomas treated with peptide-based vaccination.

Project description:Background: Pseudoprogression (PsP) mimics true early progression (TeP) in conventional imaging, which poses a diagnostic challenge in glioblastoma (GBM) patients who undergo standard concurrent chemoradiation (CCRT). This study aimed to investigate whether perioperative markers could distinguish and predict PsP from TeP in de novo isocitrate dehydrogenase (IDH) wild-type GBM patients. Methods: New or progressive gadolinium-enhancing lesions that emerged within 12 weeks after CCRT were defined as early progression. Lesions that remained stable or spontaneously regressed were classified as PsP, otherwise persistently enlarged as TeP. Clinical, radiological, and molecular information were collected for further analysis. Patients in the early progression subgroup were divided into derivation and validation sets (7:3, according to operation date). Results: Among 234 consecutive cases enrolled in this retrospective study, the incidences of PsP, TeP, and neither patterns of progression (nP) were 26.1% (61/234), 37.6% (88/234), and 36.3% (85/234), respectively. In the early progression subgroup, univariate analysis demonstrated female (OR: 2.161, P = 0.026), gross total removal (GTR) of the tumor (OR: 6.571, P < 001), located in the frontal lobe (OR: 2.561, P = 0.008), non-subventricular zone (SVZ) infringement (OR: 10.937, P < 0.001), and methylated O-6-methylguanine-DNA methyltransferase (MGMT) promoter (mMGMTp) (OR: 9.737, P < 0.001) were correlated with PsP, while GTR, non-SVZ infringement, and mMGMTp were further validated in multivariate analysis. Integrating quantitative MGMTp methylation levels from pyrosequencing, GTR, and non-SVZ infringement showed the best discriminative ability in the random forest model for derivation and validation set (AUC: 0.937, 0.911, respectively). Furthermore, a nomogram could effectively evaluate the importance of those markers in developing PsP (C-index: 0.916) and had a well-fitted calibration curve. Conclusion: Integrating those clinical, radiological, and molecular features provided a novel and robust method to distinguish PsP from TeP, which was crucial for subsequent clinical decision making, clinical trial enrollment, and prognostic assessment. By in-depth interrogation of perioperative markers, clinicians could distinguish PsP from TeP independent from advanced imaging.

Dataset Information

Discriminating pseudoprogression and true progression in diffuse infiltrating glioma using multi-parametric MRI data through deep learning.

Publications

Discriminating pseudoprogression and true progression in diffuse infiltrating glioma using multi-parametric MRI data through deep learning.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets