Reproducible Radiomics Features from Multi-MRI-Scanner Test-Retest-Study: Influence on Performance and Generalizability of Models.

BACKGROUND: Radiomics models trained on data from one center typically show a decline of performance when applied to data from external centers, hindering their introduction into large-scale clinical practice. Current expert recommendations suggest to use only reproducible radiomics features isolated by multiscanner test-retest experiments, which might help to overcome the problem of limited generalizability to external data. PURPOSE: To evaluate the influence of using only a subset of robust radiomics features, defined in a prior in vivo multi-MRI-scanner test-retest-study, on the performance and generalizability of radiomics models. STUDY TYPE: Retrospective. POPULATION: Patients with monoclonal plasma cell disorders. Training set (117 MRIs from center 1); internal test set (42 MRIs from center 1); external test set (143 MRIs from center 2-8). FIELD STRENGTH/SEQUENCE: 1.5T and 3.0T; T1-weighted turbo spin echo. ASSESSMENT: The task for the radiomics models was to predict plasma cell infiltration, determined by bone marrow biopsy, noninvasively from MRI. Radiomics machine learning models, including linear regressor, support vector regressor (SVR), and random forest regressor (RFR), were trained on data from center 1, using either all radiomics features, or using only reproducible radiomics features. Models were tested on an internal (center 1) and a multicentric external data set (center 2-8). STATISTICAL TESTS: Pearson correlation coefficient r and mean absolute error (MAE) between predicted and actual plasma cell infiltration. Fisher's z-transformation, Wilcoxon signed-rank test, Wilcoxon rank-sum test; significance level P < 0.05. RESULTS: When using only reproducible features compared with all features, the performance of the SVR on the external test set significantly improved (r = 0.43 vs. r = 0.18 and MAE = 22.6 vs. MAE = 28.2). For the RFR, the performance on the external test set deteriorated when using only reproducible instead of all radiomics features (r = 0.33 vs. r = 0.44, P = 0.29 and MAE = 21.9 vs. MAE = 20.5, P = 0.10). CONCLUSION: Using only reproducible radiomics features improves the external performance of some, but not all machine learning models, and did not automatically lead to an improvement of the external performance of the overall best radiomics model. TECHNICAL EFFICACY: Stage 2.

Correlation of visual field deficits and q-ball high-resolution fiber tractography of the optic radiation for adjacently located intracerebral lesions: preliminary results from a single-center prospective study.

Visual field deficits (VFDs) are common in patients with temporal and occipital lobe lesions. Diffusion tensor fiber tractography (DTI-FT) is widely used for surgery planning to reduce VFDs. Q-ball high-resolution fiber tractography (QBI-HRFT) improves upon DTI. This study aims to evaluate the effectiveness of DTI-FT and QBI-HRFT for surgery planning near the optic radiation (OR) as well as the correlation between VFDs, the nearest distance from the lesion to the OR fiber bundle (nD-LOR), and the lesion volume (LV). This ongoing prospective clinical trial collects clinical and imaging data of patients with lesions in deterrent areas. The present subanalysis included eight patients with gliomas near the OR. Probabilistic HRFT based on QBI-FT and conventional DTI-FT were performed for OR reconstruction based on a standard diffusion-weighted magnetic resonance imaging sequence in clinical use. Quantitative analysis was used to evaluate the lesion volume (LV) and nD-LOR. VFDs were determined based on standardized automated perimetry. We included eight patients (mean age 51.7 years [standard deviation (SD) 9.5]) with lesions near the OR. Among them, five, two, and one patients had temporodorsal, occipital, and temporal lesions, respectively. Four patients had normal vision preoperatively, while four patients had preexisting VFD. QBI-FT analysis indicated that patients with VFD exhibited a significantly smaller median nD-LOR (mean, -4.5; range -7.0; -2.3) than patients without VFD (mean, 7.4; range -4.3; 27.2) (p = 0.050). There was a trend towards a correlation between tumor volume and nD-LOR when QBI-FT was used (rs = -0.6; p = 0.056). A meticulous classification of the spatial relationship between the lesions and OR according to DTI-FT and QBI-FT was performed. The results indicated that the most prevalent orientations were the FT bundles located laterally and intrinsically in relation to the tumor. Compared with conventional DTI-FT, QBI-FT suggests reliable and more accurate results when correlated to preoperative VFDs and might be preferred for preoperative planning and intraoperative use of nearby lesions, particularly for those with larger volumes. A detailed analysis of localization, surgical approach together with QBI-FT and DTI-FT could reduce postoperative morbidity regarding VFDs. The display of HRFT techniques intraoperatively within the navigation system should be pursued for this issue.

Radiomic tractometry reveals tract-specific imaging biomarkers in white matter.

Tract-specific microstructural analysis of the brain's white matter (WM) using diffusion MRI has been a driver for neuroscientific discovery with a wide range of applications. Tractometry enables localized tissue analysis along tracts but relies on bare summary statistics and reduces complex image information along a tract to few scalar values, and so may miss valuable information. This hampers the applicability of tractometry for predictive modelling. Radiomics is a promising method based on the analysis of numerous quantitative image features beyond what can be visually perceived, but has not yet been used for tract-specific analysis of white matter. Here we introduce radiomic tractometry (RadTract) and show that introducing rich radiomics-based feature sets into the world of tractometry enables improved predictive modelling while retaining the localization capability of tractometry. We demonstrate its value in a series of clinical populations, showcasing its performance in diagnosing disease subgroups in different datasets, as well as estimation of demographic and clinical parameters. We propose that RadTract could spark the establishment of a new generation of tract-specific imaging biomarkers with benefits for a range of applications from basic neuroscience to medical research.

Deep learning aided preoperative diagnosis of primary central nervous system lymphoma.

The preoperative distinction between glioblastoma (GBM) and primary central nervous system lymphoma (PCNSL) can be difficult, even for experts, but is highly relevant. We aimed to develop an easy-to-use algorithm, based on a convolutional neural network (CNN) to preoperatively discern PCNSL from GBM and systematically compare its performance to experienced neurosurgeons and radiologists. To this end, a CNN-based on DenseNet169 was trained with the magnetic resonance (MR)-imaging data of 68 PCNSL and 69 GBM patients and its performance compared to six trained experts on an external test set of 10 PCNSL and 10 GBM. Our neural network predicted PCNSL with an accuracy of 80% and a negative predictive value (NPV) of 0.8, exceeding the accuracy achieved by clinicians (73%, NPV 0.77). Combining expert rating with automated diagnosis in those cases where experts dissented yielded an accuracy of 95%. Our approach has the potential to significantly augment the preoperative radiological diagnosis of PCNSL.