Dataset Information

Comparing effectiveness of image perturbation and test retest imaging in improving radiomic model reliability.

ABSTRACT: Image perturbation is a promising technique to assess radiomic feature repeatability, but whether it can achieve the same effect as test-retest imaging on model reliability is unknown. This study aimed to compare radiomic model reliability based on repeatable features determined by the two methods using four different classifiers. A 191-patient public breast cancer dataset with 71 test-retest scans was used with pre-determined 117 training and 74 testing samples. We collected apparent diffusion coefficient images and manual tumor segmentations for radiomic feature extraction. Random translations, rotations, and contour randomizations were performed on the training images, and intra-class correlation coefficient (ICC) was used to filter high repeatable features. We evaluated model reliability in both internal generalizability and robustness, which were quantified by training and testing AUC and prediction ICC. Higher testing performance was found at higher feature ICC thresholds, but it dropped significantly at ICC = 0.95 for the test-retest model. Similar optimal reliability can be achieved with testing AUC = 0.7-0.8 and prediction ICC > 0.9 at the ICC threshold of 0.9. It is recommended to include feature repeatability analysis using image perturbation in any radiomic study when test-retest is not feasible, but care should be taken when deciding the optimal feature repeatability criteria.

SUBMITTER: Zhang J

PROVIDER: S-EPMC10600245 | biostudies-literature | 2023 Oct

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Comparing effectiveness of image perturbation and test retest imaging in improving radiomic model reliability.

Zhang Jiang J Teng Xinzhi X Zhang Xinyu X Lam Sai-Kit SK Lin Zhongshi Z Liang Yongyi Y Yu Hao H Siu Steven Wai Kwan SWK Chang Amy Tien Yee ATY Zhang Hua H Kong Feng-Ming FM Yang Ruijie R Cai Jing J

Scientific reports 20231025 1

Image perturbation is a promising technique to assess radiomic feature repeatability, but whether it can achieve the same effect as test-retest imaging on model reliability is unknown. This study aimed to compare radiomic model reliability based on repeatable features determined by the two methods using four different classifiers. A 191-patient public breast cancer dataset with 71 test-retest scans was used with pre-determined 117 training and 74 testing samples. We collected apparent diffusion ...[more]

PMID: 37880324

Similar Datasets

Project description:ObjectiveTo examine the feasibility and test-retest reliability of encoding-task functional magnetic resonance imaging (fMRI) in mild Alzheimer disease (AD).DesignRandomized, double-blind, placebo-controlled study.SettingMemory clinical trials unit.ParticipantsWe studied 12 patients with mild AD (mean [SEM] Mini-Mental State Examination score, 24.0 [0.7]; mean Clinical Dementia Rating score, 1.0) who had been taking donepezil hydrochloride for more than 6 months from the placebo arm of a larger 24-week study (n = 24, 4 scans on weeks 0, 6, 12, and 24, respectively).InterventionsPlacebo and 3 face-name, paired-associate encoding, block-design blood oxygenation level-dependent fMRI scans in 12 weeks.Main outcome measuresWe performed whole-brain t maps (P < .001, 5 contiguous voxels) and hippocampal regions-of-interest analyses of extent (percentage of active voxels) and magnitude (percentage of signal change) for novel-greater-than-repeated face-name contrasts. We also calculated intraclass correlation coefficients and power estimates for hippocampal regions of interest.ResultsTask tolerability and data yield were high (95 of 96 scans yielded favorable-quality data). Whole-brain maps were stable. Right and left hippocampal regions-of-interest intraclass correlation coefficients were 0.59 to 0.87 and 0.67 to 0.74, respectively. To detect 25.0% to 50.0% changes in week-0 to week-12 hippocampal activity using left-right extent or right magnitude with 80.0% power (2-sided α = .05) requires 14 to 51 patients. Using left magnitude requires 125 patients because of relatively small signal to variance ratios.ConclusionsEncoding-task fMRI was successfully implemented in a single-site, 24-week, AD randomized controlled trial. Week 0 to 12 whole-brain t maps were stable, and test-retest reliability of hippocampal fMRI measures ranged from moderate to substantial. Right hippocampal magnitude may be the most promising of these candidate measures in a leveraged context. These initial estimates of test-retest reliability and power justify evaluation of encoding-task fMRI as a potential biomarker for signal of effect in exploratory and proof-of-concept trials in mild AD. Validation of these results with larger sample sizes and assessment in multisite studies is warranted.

Dataset Information

Comparing effectiveness of image perturbation and test retest imaging in improving radiomic model reliability.

Publications

Comparing effectiveness of image perturbation and test retest imaging in improving radiomic model reliability.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets