Evaluating intensity normalization on MRIs of human brain with multiple sclerosis

MR-based radiomics predictive modelling of EGFR mutation and HER2 overexpression in metastatic brain adenocarcinoma: a two-centre study

OBJECTIVES

Magnetic resonance (MR)-based radiomics features of brain metastases are utilised to predict epidermal growth factor receptor (EGFR) mutation and human epidermal growth factor receptor 2 (HER2) overexpression in adenocarcinoma, with the aim to identify the most predictive MR sequence.

METHODS

A retrospective inclusion of 268 individuals with brain metastases from adenocarcinoma across two institutions was conducted. Utilising T1-weighted imaging (T1 contrast-enhanced [T1-CE]) and T2 fluid-attenuated inversion recovery (T2-FLAIR) sequences, 1,409 radiomics features were extracted. These sequences were randomly divided into training and test sets at a 7:3 ratio. The selection of relevant features was done using the least absolute shrinkage selection operator, and the training cohort's support vector classifier model was employed to generate the predictive model. The performance of the radiomics features was evaluated using a separate test set.

RESULTS

For contrast-enhanced T1-CE cohorts, the radiomics features based on 19 selected characteristics exhibited excellent discrimination. No significant differences in age, sex, and time to metastasis were observed between the groups with EGFR mutations or HER2 + and those with wild-type EGFR or HER2 (p > 0.05). Radiomics feature analysis for T1-CE revealed an area under the curve (AUC) of 0.98, classification accuracy of 0.93, sensitivity of 0.92, and specificity of 0.93 in the training cohort. In the test set, the AUC was 0.82. The 19 radiomics features for the T2-FLAIR sequence showed AUCs of 0.86 in the training set and 0.70 in the test set.

CONCLUSIONS

This study developed a T1-CE signature that could serve as a non-invasive adjunctive tool to determine the presence of EGFR mutations and HER2 + status in adenocarcinoma, aiding in the direction of treatment plans.

CLINICAL RELEVANCE STATEMENT

We propose radiomics features based on T1-CE brain MR sequences that are both evidence-based and non-invasive. These can be employed to guide clinical treatment planning in patients with brain metastases from adenocarcinoma.

Longitudinal changes in pigment epithelial detachment composition indices (PEDCI): new biomarkers in neovascular age-related macular degeneration

PURPOSE

To evaluate novel, automated biomarkers, pigment epithelial detachment composition indices (PEDCI) in eyes with neovascular age-related macular degeneration (nAMD) undergoing anti-vascular endothelial growth factor (anti-VEGF) therapy through 24 months.

METHODS

Retrospective analysis of 37 eyes (34 patients) with PED associated with nAMD receiving as-needed anti-VEGF treatment was performed. Best-corrected visual acuity (BCVA) and optical coherence tomography images were acquired at a treatment-naïve baseline and 3-, 6-, 12-, 18-, and 24-month visits. Previously validated automated imaging biomarkers, PEDCI-S (serous), PEDCI-N (neovascular), and PEDCI-F (fibrous) within PEDs were measured. ANOVA analysis and Spearman correlation were performed.

RESULTS

Mean BCVA (in logMAR) was 0.60 ± 0.47, 0.45 ± 0.41, 0.49 ± 0.49, 0.61 ± 0.54, 0.59 ± 0.56, and 0.67 ± 0.57 at baseline, 3, 6, 12, 18, and 24 months respectively. Overall, BCVA showed minimal worsening of 0.07 ± 0.54 logMAR (p = 0.07). 13.38 ± 3.77 anti-VEGF injections were given through 24 months. PEDCI-F showed an increase of 0.116, 0.122, 0.036, and 0.006 at months 3, 6, 12, and 18 respectively and a decrease of 0.004 at month 24 (p = 0.03); PEDCI-S showed a decrease of 0.064, 0.130, 0.091, 0.092, and 0.095 at months 3, 6, 12, 18, and 24 respectively (p = 0.16); PEDCI-N showed a decrease of 0.052 at month 3 and an increase of 0.008, 0.055, 0.086, and 0.099 at months 6, 12, 18, and 24 respectively (p = 0.06). BCVA was negatively correlated with PEDCI-F (r = -0.28, p < 0.01), and positively correlated with PEDCI-N (r = 0.28, p < 0.01) and PEDCI-S (r = 0.15, p = 0.03).

CONCLUSION

Longitudinal analysis of PEDCI supports their utility as biomarkers that characterize treatment related effects by quantifying the relative composition of PEDs.

MRI-based radiomics for predicting histology in malignant salivary gland tumors: methodology and "proof of principle"

Defining the exact histological features of salivary gland malignancies before treatment remains an unsolved problem that compromises the ability to tailor further therapeutic steps individually. Radiomics, a new methodology to extract quantitative information from medical images, could contribute to characterizing the individual cancer phenotype already before treatment in a fast and non-invasive way. Consequently, the standardization and implementation of radiomic analysis in the clinical routine work to predict histology of salivary gland cancer (SGC) could also provide improvements in clinical decision-making. In this study, we aimed to investigate the potential of radiomic features as imaging biomarker to distinguish between high grade and low-grade salivary gland malignancies. We have also investigated the effect of image and feature level harmonization on the performance of radiomic models. For this study, our dual center cohort consisted of 126 patients, with histologically proven SGC, who underwent curative-intent treatment in two tertiary oncology centers. We extracted and analyzed the radiomics features of 120 pre-therapeutic MRI images with gadolinium (T1 sequences), and correlated those with the definitive post-operative histology. In our study the best radiomic model achieved average AUC of 0.66 and balanced accuracy of 0.63. According to the results, there is significant difference between the performance of models based on MRI intensity normalized images + harmonized features and other models (p value < 0.05) which indicates that in case of dealing with heterogeneous dataset, applying the harmonization methods is beneficial. Among radiomic features minimum intensity from first order, and gray level-variance from texture category were frequently selected during multivariate analysis which indicate the potential of these features as being used as imaging biomarker. The present bicentric study presents for the first time the feasibility of implementing MR-based, handcrafted radiomics, based on T1 contrast-enhanced sequences and the ComBat harmonization method in an effort to predict the formal grading of salivary gland carcinoma with satisfactory performance.

Magnetic resonance imaging radiomic features stability in brain metastases: Impact of image preprocessing, image-, and feature-level harmonization

Background and purpose

Magnetic resonance imaging (MRI) scans are highly sensitive to acquisition and reconstruction parameters which affect feature stability and model generalizability in radiomic research. This work aims to investigate the effect of image pre-processing and harmonization methods on the stability of brain MRI radiomic features and the prediction performance of radiomic models in patients with brain metastases (BMs).

Materials and methods

Two T1 contrast enhanced brain MRI data-sets were used in this study. The first contained 25 BMs patients with scans at two different time points and was used for features stability analysis. The effect of gray level discretization (GLD), intensity normalization (Z-score, Nyul, WhiteStripe, and in house-developed method named N-Peaks), and ComBat harmonization on features stability was investigated and features with intraclass correlation coefficient >0.8 were considered as stable. The second data-set containing 64 BMs patients was used for a classification task to investigate the informativeness of stable features and the effects of harmonization methods on radiomic model performance.

Results

Applying fixed bin number (FBN) GLD, resulted in higher number of stable features compare to fixed bin size (FBS) discretization (10 ± 5.5 % higher). `Harmonization in feature domain improved the stability for non-normalized and normalized images with Z-score and WhiteStripe methods. For the classification task, keeping the stable features resulted in good performance only for normalized images with N-Peaks along with FBS discretization.

Conclusions

To develop a robust MRI based radiomic model we recommend using an intensity normalization method based on a reference tissue (e.g N-Peaks) and then using FBS discretization.

Learning multi-site harmonization of magnetic resonance images without traveling human phantoms

Harmonization improves Magn. Reson. Imaging (MRI) data consistency and is central to effective integration of diverse imaging data acquired across multiple sites. Recent deep learning techniques for harmonization are predominantly supervised in nature and hence require imaging data of the same human subjects to be acquired at multiple sites. Data collection as such requires the human subjects to travel across sites and is hence challenging, costly, and impractical, more so when sufficient sample size is needed for reliable network training. Here we show how harmonization can be achieved with a deep neural network that does not rely on traveling human phantom data. Our method disentangles site-specific appearance information and site-invariant anatomical information from images acquired at multiple sites and then employs the disentangled information to generate the image of each subject for any target site. We demonstrate with more than 6,000 multi-site T1- and T2-weighted images that our method is remarkably effective in generating images with realistic site-specific appearances without altering anatomical details. Our method allows retrospective harmonization of data in a wide range of existing modern large-scale imaging studies, conducted via different scanners and protocols, without additional data collection.

High resolution 0.5mm isotropic T1-weighted and diffusion tensor templates of the brain of non-demented older adults in a common space for the MIITRA atlas

High quality, high resolution T1-weighted (T1w) and diffusion tensor imaging (DTI) brain templates located in a common space can enhance the sensitivity and precision of template-based neuroimaging studies. However, such multimodal templates have not been constructed for the older adult brain. The purpose of this work which is part of the MIITRA atlas project was twofold: (A) to develop 0.5 mm isotropic resolution T1w and DTI templates that are representative of the brain of non-demented older adults and are located in the same space, using advanced multimodal template construction techniques and principles of super resolution on data from a large, diverse, community cohort of 400 non-demented older adults, and (B) to systematically compare the new templates to other standardized templates. It was demonstrated that the new MIITRA-0.5mm T1w and DTI templates are well-matched in space, exhibit good definition of brain structures, including fine structures, exhibit higher image sharpness than other standardized templates, and are free of artifacts. The MIITRA-0.5mm T1w and DTI templates allowed higher intra-modality inter-subject spatial normalization precision as well as higher inter-modality intra-subject spatial matching of older adult T1w and DTI data compared to other available templates. Consequently, MIITRA-0.5mm templates allowed detection of smaller inter-group differences for older adult data compared to other templates. The MIITRA-0.5mm templates were also shown to be most representative of the brain of non-demented older adults compared to other templates with submillimeter resolution. The new templates constructed in this work constitute two of the final products of the MIITRA atlas project and are anticipated to have important implications for the sensitivity and precision of studies on older adults.

Artificial Cognition for Detection of Mental Disability: A Vision Transformer Approach for Alzheimer's Disease

Alzheimer's disease is a common neurological disorder and mental disability that causes memory loss and cognitive decline, presenting a major challenge to public health due to its impact on millions of individuals worldwide. It is crucial to diagnose and treat Alzheimer's in a timely manner to improve the quality of life of both patients and caregivers. In the recent past, machine learning techniques have showed potential in detecting Alzheimer's disease by examining neuroimaging data, especially Magnetic Resonance Imaging (MRI). This research proposes an attention-based mechanism that employs the vision transformer approach to detect Alzheimer's using MRI images. The presented technique applies preprocessing to the MRI images and forwards them to a vision transformer network for classification. This network is trained on the publicly available Kaggle dataset, and it illustrated impressive results with an accuracy of 99.06%, precision of 99.06%, recall of 99.14%, and F1-score of 99.1%. Furthermore, a comparative study is also conducted to evaluate the performance of the proposed method against various state-of-the-art techniques on diverse datasets. The proposed method demonstrated superior performance, outperforming other published methods when applied to the Kaggle dataset.

Association of graph-based spatial features with overall survival status of glioblastoma patients

Glioblastoma is the most common malignant brain tumor with less than 15 months median survival. To aid prognosis, there is a need for decision tools that leverage diagnostic modalities such as MRI to inform survival. In this study, we examine higher-order spatial proximity characteristics from habitats and propose two graph-based methods (minimum spanning tree and graph run-length matrix) to characterize spatial heterogeneity over tumor MRI-derived intensity habitats and assess their relationships with overall survival as well as the immune signature status of patients with glioblastoma. A data set of 74 patients was studied based on the availability of post-contrast T1-weighted and T2-weighted fluid attenuated inversion recovery (FLAIR) image data in The Cancer Image Archive (TCIA). We assessed the predictive value of MST- and GRLM-derived features from 2D images for prediction of 12-month survival status and immune signature status of patients with glioblastoma via a receiver operating characteristic curve analysis. For 12-month survival prediction using MST-based method, sensitivity and specificity were 0.82 and 0.79 respectively. For GRLM-based method, sensitivity and specificity were 0.73 and 0.77 respectively. For immune status, sensitivity and specificity were 0.91 and 0.69, respectively, for the GRLM-based method with an immune effector. Our results show that the proposed MST- and GRLM-derived features are predictive of 12-month survival status as well as the immune signature status of patients with glioblastoma. To our knowledge, this is the first application of MST- and GRLM-based proximity analyses for the study of radiologically-defined tumor habitats in glioblastoma.

Radiomics-Based Prediction of Anti-VEGF Treatment Response in Neovascular Age-Related Macular Degeneration With Pigment Epithelial Detachment

Purpose

Machine learning models based on radiomic feature extraction from clinical imaging data provide effective and interpretable means for clinical decision making. This pilot study evaluated whether radiomics features in baseline optical coherence tomography (OCT) images of eyes with pigment epithelial detachment (PED) associated with neovascular age-related macular degeneration (nAMD) can predict treatment response to as-needed anti-vascular endothelial growth factor (VEGF) therapy.

Methods

Thirty-nine eyes of patients with PED undergoing anti-VEGF therapy were included. All eyes underwent a loading dose followed by as-needed therapy. OCT images at baseline, month 3, and month 6 were analyzed. Images were manually separated into non-responding, recurring, and responding eyes based on the presence or absence of subretinal fluid at month 6. PED radiomics features were then extracted from each image and images were classified as responding or recurring using a machine learning classifier applied to the radiomics features.

Results

Linear discriminant analysis classification of baseline features as responsive versus recurring resulted in classification performance of 64.0% (95% confidence interval [CI] = 0.63-0.65), area under the curve (AUC = 0.78, 95% CI = 0.72-0.82), sensitivity 0.79 (95% CI = 0.63-0.87), and specificity 0.58 (95% CI = 0.50-0.67). Further analysis of features in recurring eyes identified a significant shift toward non-responding mean feature values over 6 months.

Conclusions

Our results demonstrate the use of radiomics features as predictors for treatment response to as-needed anti-VEGF therapy. Our study demonstrates the potential for radiomics feature in clinical decision support for personalizing anti-VEGF therapy.

Translational Relevance

The ability to use PED texture features to predict treatment response facilitates personalized clinical decision making.

MISPEL: A supervised deep learning harmonization method for multi-scanner neuroimaging data

Large-scale data obtained from aggregation of already collected multi-site neuroimaging datasets has brought benefits such as higher statistical power, reliability, and robustness to the studies. Despite these promises from growth in sample size, substantial technical variability stemming from differences in scanner specifications exists in the aggregated data and could inadvertently bias any downstream analyses on it. Such a challenge calls for data normalization and/or harmonization frameworks, in addition to comprehensive criteria to estimate the scanner-related variability and evaluate the harmonization frameworks. In this study, we propose MISPEL (Multi-scanner Image harmonization via Structure Preserving Embedding Learning), a supervised multi-scanner harmonization method that is naturally extendable to more than two scanners. We also designed a set of criteria to investigate the scanner-related technical variability and evaluate the harmonization techniques. As an essential requirement of our criteria, we introduced a multi-scanner matched dataset of 3T T1 images across four scanners, which, to the best of our knowledge is one of the few datasets of this kind. We also investigated our evaluations using two popular segmentation frameworks: FSL and segmentation in statistical parametric mapping (SPM). Lastly, we compared MISPEL to popular methods of normalization and harmonization, namely White Stripe, RAVEL, and CALAMITI. MISPEL outperformed these methods and is promising for many other neuroimaging modalities.

Individual Variability of Human Cortical Spontaneous Activity by 3T/7T fMRI

Mapping variability in cortical spontaneous activity (CSA) is an essential goal of understanding various sources of dark brain energy in human neuroscience. CSA was traditionally characterized using resting-state functional MRI (rfMRI) at 1.5T or 3T magnets while recently with 7T-rfMRI. However, the utility and interpretability of 7T-rfMRI must first be established for its variability. By leveraging rfMRI data from the Human Connectome Project (HCP), we derived CSA metrics with 3T-rfMRI and 7T-rfMRI for the same 84 healthy participants (52 females). The 7T-rfMRI produces different CSA metrics at multiple spatial-scales and their variability from the 3T-rfMRI. These differences were spatially dependent and varied according to specific cortical organization. For the amplitude metric, 7T-rfMRI enhanced its spatial contrasts in the anterior cortex but weakened it in the posterior cortex. An opposite pattern was observed for the connectivity metrics. The reliability changes of these metrics were scale dependent, indicating enhanced reliability for connectivity but weakened reliability for amplitude by 7T-rfMRI. These effects were primarily located in the high-order associate cortex, parsing the corresponding changes in individual differences with respect to 7T-rfMRI: (1) higher connectivity variability between participants and the lower connectivity variability within individual participants, and (2) lower amplitude variability between participants and higher amplitude variability within participants. Our work, for the first time, demonstrated the variability of the human CSA across space, rfMRI settings/platforms, and individuals. We discussed the statistical implications of our findings on CSA-based experimental designs and reproducible neuroscience as well as their translational value for personalized applications.

Bayesian statistical modeling to predict observer-specific optimal windowing parameters in magnetic resonance imaging

Magnetic resonance (MR) images require a process known as windowing for optimizing the display conditions. However, the conventional windowing process often fails to achieve the preferred display conditions for observers due to various factors. This study proposes a novel framework for predicting the preferred windowing parameters for each observer using Bayesian statistical modeling. MR images obtained from 1000 patients were divided into training and test sets at a 7:3 ratio. The image intensity and windowing parameters were standardized using previously reported methods. Bayesian statistical modeling was utilized to predict the windowing parameters preferred by three MR imaging (MRI) operators. The performance of the proposed framework was evaluated by assessing the mean relative error (MRE), mean absolute error (MAE), and Pearson's correlation coefficient (ρ) of the test set. In addition, the naive method, which presumes that the average value of the windowing parameters for each acquisition sequence and body region in the training set is optimal, was also used for comparison. Three MRI operators and three radiologists conducted visual assessments. The mean MRE, MAE, and ρ values for the window level and width (WL/WW) in the proposed framework were 12.6 and 13.9, 42.9 and 85.4, and 0.98 and 0.98, respectively. These results outperformed those obtained using the naive method. The visual assessments revealed no significant differences between the original and predicted display conditions, indicating that the proposed framework accurately predicts individualized windowing parameters with the additional advantages of robustness and ease of use. Thus, the proposed framework can effectively predict the windowing parameters preferred by each observer.

MR Intensity Normalization Methods Impact Sequence Specific Radiomics Prognostic Model Performance in Primary and Recurrent High-Grade Glioma

PURPOSE

This study investigates the impact of different intensity normalization (IN) methods on the overall survival (OS) radiomics models' performance of MR sequences in primary (pHGG) and recurrent high-grade glioma (rHGG).

METHODS

MR scans acquired before radiotherapy were retrieved from two independent cohorts (rHGG C1: 197, pHGG C2: 141) from multiple scanners (15, 14). The sequences are T1 weighted (w), contrast-enhanced T1w (T1wce), T2w, and T2w-FLAIR. Sequence-specific significant features (SF) associated with OS, extracted from the tumour volume, were derived after applying 15 different IN methods. Survival analyses were conducted using Cox proportional hazard (CPH) and Poisson regression (POI) models. A ranking score was assigned based on the 10-fold cross-validated (CV) concordance index (C-I), mean square error (MSE), and the Akaike information criterion (AICs), to evaluate the methods' performance.

RESULTS

Scatter plots of the 10-CV C-I and MSE against the AIC showed an impact on the survival predictions between the IN methods and MR sequences (C1/C2 C-I range: 0.62-0.71/0.61-0.72, MSE range: 0.20-0.42/0.13-0.22). White stripe showed stable results for T1wce (C1/C2 C-I: 0.71/0.65, MSE: 0.21/0.14). Combat (0.68/0.62, 0.22/0.15) and histogram matching (HM, 0.67/0.64, 0.22/0.15) showed consistent prediction results for T2w models. They were also the top-performing methods for T1w in C2 (Combat: 0.67, 0.13; HM: 0.67, 0.13); however, only HM achieved high predictions in C1 (0.66, 0.22). After eliminating IN impacted SF using Spearman's rank-order correlation coefficient, a mean decrease in the C-I and MSE of 0.05 and 0.03 was observed in all four sequences.

CONCLUSION

The IN method impacted the predictive power of survival models; thus, performance is sequence-dependent.

Pigment epithelial detachment composition indices (PEDCI) in neovascular age-related macular degeneration

We provide an automated analysis of the pigment epithelial detachments (PEDs) in neovascular age-related macular degeneration (nAMD) and estimate areas of serous, neovascular, and fibrous tissues within PEDs. A retrospective analysis of high-definition spectral-domain OCT B-scans from 43 eyes of 37 patients with nAMD with presence of fibrovascular PED was done. PEDs were manually segmented and then filtered using 2D kernels to classify pixels within the PED as serous, neovascular, or fibrous. A set of PED composition indices were calculated on a per-image basis using relative PED area of serous (PEDCI-S), neovascular (PEDCI-N), and fibrous (PEDCI-F) tissue. Accuracy of segmentation and classification within the PED were graded in masked fashion. Mean overall intra-observer repeatability and inter-observer reproducibility were 0.86 ± 0.07 and 0.86 ± 0.03 respectively using intraclass correlations. The mean graded scores were 96.99 ± 8.18, 92.12 ± 7.97, 91.48 ± 8.93, and 92.29 ± 8.97 for segmentation, serous, neovascular, and fibrous respectively. Mean (range) PEDCI-S, PEDCI-N, and PEDCI-F were 0.253 (0-0.952), 0.554 (0-1), and 0.193 (0-0.693). A kernel-based image processing approach demonstrates potential for approximating PED composition. Evaluating follow up changes during nAMD treatment with respect to PEDCI would be useful for further clinical applications.

Harmonization Strategies in Multicenter MRI-Based Radiomics

Radiomics analysis is a powerful tool aiming to provide diagnostic and prognostic patient information directly from images that are decoded into handcrafted features, comprising descriptors of shape, size and textural patterns. Although radiomics is gaining momentum since it holds great promise for accelerating digital diagnostics, it is susceptible to bias and variation due to numerous inter-patient factors (e.g., patient age and gender) as well as inter-scanner ones (different protocol acquisition depending on the scanner center). A variety of image and feature based harmonization methods has been developed to compensate for these effects; however, to the best of our knowledge, none of these techniques has been established as the most effective in the analysis pipeline so far. To this end, this review provides an overview of the challenges in optimizing radiomics analysis, and a concise summary of the most relevant harmonization techniques, aiming to provide a thorough guide to the radiomics harmonization process.

Intensity standardization of MRI prior to radiomic feature extraction for artificial intelligence research in glioma-a systematic review

OBJECTIVES

Radiomics is a promising avenue in non-invasive characterisation of diffuse glioma. Clinical translation is hampered by lack of reproducibility across centres and difficulty in standardising image intensity in MRI datasets. The study aim was to perform a systematic review of different methods of MRI intensity standardisation prior to radiomic feature extraction.

METHODS

MEDLINE, EMBASE, and SCOPUS were searched for articles meeting the following eligibility criteria: MRI radiomic studies where one method of intensity normalisation was compared with another or no normalisation, and original research concerning patients diagnosed with diffuse gliomas. Using PRISMA criteria, data were extracted from short-listed studies including number of patients, MRI sequences, validation status, radiomics software, method of segmentation, and intensity standardisation. QUADAS-2 was used for quality appraisal.

RESULTS

After duplicate removal, 741 results were returned from database and reference searches and, from these, 12 papers were eligible. Due to a lack of common pre-processing and different analyses, a narrative synthesis was sought. Three different intensity standardisation techniques have been studied: histogram matching (5/12), limiting or rescaling signal intensity (8/12), and deep learning (1/12)-only two papers compared different methods. From these studies, histogram matching produced the more reliable features compared to other methods of altering MRI signal intensity.

CONCLUSION

Multiple methods of intensity standardisation have been described in the literature without clear consensus. Further research that directly compares different methods of intensity standardisation on glioma MRI datasets is required.

KEY POINTS

• Intensity standardisation is a key pre-processing step in the development of robust radiomic signatures to evaluate diffuse glioma. • A minority of studies compared the impact of two or more methods. • Further research is required to directly compare multiple methods of MRI intensity standardisation on glioma datasets.

Artificial intelligence in the radiomic analysis of glioblastomas: A review, taxonomy, and perspective

Radiological imaging techniques, including magnetic resonance imaging (MRI) and positron emission tomography (PET), are the standard-of-care non-invasive diagnostic approaches widely applied in neuro-oncology. Unfortunately, accurate interpretation of radiological imaging data is constantly challenged by the indistinguishable radiological image features shared by different pathological changes associated with tumor progression and/or various therapeutic interventions. In recent years, machine learning (ML)-based artificial intelligence (AI) technology has been widely applied in medical image processing and bioinformatics due to its advantages in implicit image feature extraction and integrative data analysis. Despite its recent rapid development, ML technology still faces many hurdles for its broader applications in neuro-oncological radiomic analysis, such as lack of large accessible standardized real patient radiomic brain tumor data of all kinds and reliable predictions on tumor response upon various treatments. Therefore, understanding ML-based AI technologies is critically important to help us address the skyrocketing demands of neuro-oncology clinical deployments. Here, we provide an overview on the latest advancements in ML techniques for brain tumor radiomic analysis, emphasizing proprietary and public dataset preparation and state-of-the-art ML models for brain tumor diagnosis, classifications (e.g., primary and secondary tumors), discriminations between treatment effects (pseudoprogression, radiation necrosis) and true progression, survival prediction, inflammation, and identification of brain tumor biomarkers. We also compare the key features of ML models in the realm of neuroradiology with ML models employed in other medical imaging fields and discuss open research challenges and directions for future work in this nascent precision medicine area.

Data harmonisation for information fusion in digital healthcare: A state-of-the-art systematic review, meta-analysis and future research directions

Removing the bias and variance of multicentre data has always been a challenge in large scale digital healthcare studies, which requires the ability to integrate clinical features extracted from data acquired by different scanners and protocols to improve stability and robustness. Previous studies have described various computational approaches to fuse single modality multicentre datasets. However, these surveys rarely focused on evaluation metrics and lacked a checklist for computational data harmonisation studies. In this systematic review, we summarise the computational data harmonisation approaches for multi-modality data in the digital healthcare field, including harmonisation strategies and evaluation metrics based on different theories. In addition, a comprehensive checklist that summarises common practices for data harmonisation studies is proposed to guide researchers to report their research findings more effectively. Last but not least, flowcharts presenting possible ways for methodology and metric selection are proposed and the limitations of different methods have been surveyed for future research.

Harmonization of multi-site MRS data with ComBat

Magnetic resonance spectroscopy (MRS) is a non-invasive neuroimaging technique used to measure brain chemistry in vivo and has been used to study the healthy brain as well as neuropathology in numerous neurological disorders. The number of multi-site studies using MRS are increasing; however, non-biological variability introduced during data collection across multiple sites, such as differences in scanner vendors and site-specific acquisition implementations for MRS, can obscure detection of biological effects of interest. ComBat is a data harmonization technique that can remove non-biological sources of variance in multisite studies. It has been validated for use with structural and functional MRI metrics but not for MRS metabolites. This study investigated the validity of using ComBat to harmonize MRS metabolites for vendor and site differences. Analyses were performed using data acquired across 20 sites and included edited MRS for GABA+ (N=218) and macromolecule-suppressed GABA data (N=209), as well as standard PRESS data to quantify NAA, creatine, choline, and glutamate (N=190). ComBat harmonization successfully mitigated vendor and site differences for all metabolites of interest. Moreover, significant associations were detected between sex and choline levels and between age and glutamate and GABA+ levels that were not detectable prior to harmonization, confirming the importance of removing site and vendor effects in multi-site data. In conclusion, ComBat harmonization can be successfully applied to MRS data in multi-site MRS studies.

A Comprehensive Survey on the Detection, Classification, and Challenges of Neurological Disorders

Simple Summary

This study represents a resourceful review article that can deliver resources on neurological diseases and their implemented classification algorithms to reveal the future direction of researchers. Researchers interested in studying neurological diseases and previously implemented techniques in this field can follow this article. Various challenges occur in detecting different stages of the disorders. A limited amount of labeled and unlabeled datasets and other limitations is represented in this article to assist them in finding out the directions. The authors’ purpose for composing this article is to make a straightforward and concrete path for researchers to quickly find the way and the scope in this field for implementing future research on neurological disease detection.

Abstract

Neurological disorders (NDs) are becoming more common, posing a concern to pregnant women, parents, healthy infants, and children. Neurological disorders arise in a wide variety of forms, each with its own set of origins, complications, and results. In recent years, the intricacy of brain functionalities has received a better understanding due to neuroimaging modalities, such as magnetic resonance imaging (MRI), magnetoencephalography (MEG), and positron emission tomography (PET), etc. With high-performance computational tools and various machine learning (ML) and deep learning (DL) methods, these modalities have discovered exciting possibilities for identifying and diagnosing neurological disorders. This study follows a computer-aided diagnosis methodology, leading to an overview of pre-processing and feature extraction techniques. The performance of existing ML and DL approaches for detecting NDs is critically reviewed and compared in this article. A comprehensive portion of this study also shows various modalities and disease-specified datasets that detect and records images, signals, and speeches, etc. Limited related works are also summarized on NDs, as this domain has significantly fewer works focused on disease and detection criteria. Some of the standard evaluation metrics are also presented in this study for better result analysis and comparison. This research has also been outlined in a consistent workflow. At the conclusion, a mandatory discussion section has been included to elaborate on open research challenges and directions for future work in this emerging field.

Deep Generative Medical Image Harmonization for Improving Cross-Site Generalization in Deep Learning Predictors

BACKGROUND

In the medical imaging domain, deep learning-based methods have yet to see widespread clinical adoption, in part due to limited generalization performance across different imaging devices and acquisition protocols. The deviation between estimated brain age and biological age is an established biomarker of brain health and such models may benefit from increased cross-site generalizability.

PURPOSE

To develop and evaluate a deep learning-based image harmonization method to improve cross-site generalizability of deep learning age prediction.

STUDY TYPE

Retrospective.

POPULATION

Eight thousand eight hundred and seventy-six subjects from six sites. Harmonization models were trained using all subjects. Age prediction models were trained using 2739 subjects from a single site and tested using the remaining 6137 subjects from various other sites.

FIELD STRENGTH/SEQUENCE

Brain imaging with magnetization prepared rapid acquisition with gradient echo or spoiled gradient echo sequences at 1.5 T and 3 T.

ASSESSMENT

StarGAN v2, was used to perform a canonical mapping from diverse datasets to a reference domain to reduce site-based variation while preserving semantic information. Generalization performance of deep learning age prediction was evaluated using harmonized, histogram matched, and unharmonized data.

STATISTICAL TESTS

Mean absolute error (MAE) and Pearson correlation between estimated age and biological age quantified the performance of the age prediction model.

RESULTS

Our results indicated a substantial improvement in age prediction in out-of-sample data, with the overall MAE improving from 15.81 (±0.21) years to 11.86 (±0.11) with histogram matching to 7.21 (±0.22) years with generative adversarial network (GAN)-based harmonization. In the multisite case, across the 5 out-of-sample sites, MAE improved from 9.78 (±6.69) years to 7.74 (±3.03) years with histogram normalization to 5.32 (±4.07) years with GAN-based harmonization.

DATA CONCLUSION

While further research is needed, GAN-based medical image harmonization appears to be a promising tool for improving cross-site deep learning generalization.

LEVEL OF EVIDENCE

4 TECHNICAL EFFICACY: Stage 1.

Multisite reproducibility and test-retest reliability of the T1w/T2w-ratio: A comparison of processing methods

BACKGROUND

The ratio of T1-weighted (T1w) and T2-weighted (T2w) magnetic resonance imaging (MRI) images is often used as a proxy measure of cortical myelin. However, the T1w/T2w-ratio is based on signal intensities that are inherently non-quantitative and known to be affected by extrinsic factors. To account for this a variety of processing methods have been proposed, but a systematic evaluation of their efficacy is lacking. Given the dependence of the T1w/T2w-ratio on scanner hardware and T1w and T2w protocols, it is important to ensure that processing pipelines perform well also across different sites.

METHODS

We assessed a variety of processing methods for computing cortical T1w/T2w-ratio maps, including correction methods for nonlinear field inhomogeneities, local outliers, and partial volume effects as well as intensity normalisation. These were implemented in 33 processing pipelines which were applied to four test-retest datasets, with a total of 170 pairs of T1w and T2w images acquired on four different MRI scanners. We assessed processing pipelines across datasets in terms of their reproducibility of expected regional distributions of cortical myelin, lateral intensity biases, and test-retest reliability regionally and across the cortex. Regional distributions were compared both qualitatively with histology and quantitatively with two reference datasets, YA-BC and YA-B1+, from the Human Connectome Project.

RESULTS

Reproducibility of raw T1w/T2w-ratio distributions was overall high with the exception of one dataset. For this dataset, Spearman rank correlations increased from 0.27 to 0.70 after N3 bias correction relative to the YA-BC reference and from -0.04 to 0.66 after N4ITK bias correction relative to the YA-B1+ reference. Partial volume and outlier corrections had only marginal effects on the reproducibility of T1w/T2w-ratio maps and test-retest reliability. Before intensity normalisation, we found large coefficients of variation (CVs) and low intraclass correlation coefficients (ICCs), with total whole-cortex CV of 10.13% and whole-cortex ICC of 0.58 for the raw T1w/T2w-ratio. Intensity normalisation with WhiteStripe, RAVEL, and Z-Score improved total whole-cortex CVs to 5.91%, 5.68%, and 5.19% respectively, whereas Z-Score and Least Squares improved whole-cortex ICCs to 0.96 and 0.97 respectively.

CONCLUSIONS

In the presence of large intensity nonuniformities, bias field correction is necessary to achieve acceptable correspondence with known distributions of cortical myelin, but it can be detrimental in datasets with less intensity inhomogeneity. Intensity normalisation can improve test-retest reliability and inter-subject comparability. However, both bias field correction and intensity normalisation methods vary greatly in their efficacy and may affect the interpretation of results. The choice of T1w/T2w-ratio processing method must therefore be informed by both scanner and acquisition protocol as well as the given study objective. Our results highlight limitations of the T1w/T2w-ratio, but also suggest concrete ways to enhance its usefulness in future studies.

A multi-scanner neuroimaging data harmonization using RAVEL and ComBat

Modern neuroimaging studies frequently combine data collected from multiple scanners and experimental conditions. Such data often contain substantial technical variability associated with image intensity scale (image intensity scales are not the same in different images) and scanner effects (images obtained from different scanners contain substantial technical biases). Here we evaluate and compare results of data analysis methods without any data transformation (RAW), with intensity normalization using RAVEL, with regional harmonization methods using ComBat, and a combination of RAVEL and ComBat. Methods are evaluated on a unique sample of 16 study participants who were scanned on both 1.5T and 3T scanners a few months apart. Neuroradiological evaluation was conducted for 7 different regions of interest (ROI’s) pertinent to Alzheimer’s disease (AD). Cortical measures and results indicate that: (1) RAVEL substantially improved the reproducibility of image intensities; (2) ComBat is preferred over RAVEL and the RAVEL-ComBat combination in terms of regional level harmonization due to more consistent harmonization across subjects and image-derived measures; (3) RAVEL and ComBat substantially reduced bias compared to analysis of RAW images, but RAVEL also resulted in larger variance; and (4) the larger root mean square deviation (RMSD) of RAVEL compared to ComBat is due mainly to its larger variance.

Standardizing T1-w/T2-w ratio images in trigeminal neuralgia to estimate the degree of demyelination in vivo

•

The T1-w/T2-w ratio image or “myelin-sensitive map (MM)” is a non-invasive tool that can estimate myelin content in different regions of the brain and between different patients in vivo.

•

T1-w and T2-w images are standardized post-hoc using histogram matching algorithms to provide tissue-specific intensity information and facilitate MM analysis.

•

Analysis of MM intensities demonstrate reduced myelin content in MS plaques compared to its corresponding pontine regions in CTN patients and its surrounding NAWM in MSTN patients.

•

MM has the potential to distinguish changes in myelin of NAWM before MS plaques are detectable on conventional MR images.

Background

Novel magnetic resonance (MR) imaging techniques have led to the development of T1-w/T2-w ratio images or “myelin-sensitive maps (MMs)” to estimate and compare myelin content in vivo. Currently, raw image intensities in conventional MR images are unstandardized, preventing meaningful quantitative comparisons. We propose an improved workflow to standardize the MMs, which was applied to patients with classic trigeminal neuralgia (CTN) and trigeminal neuralgia secondary to multiple sclerosis (MSTN), to assess the validity and feasibility of this clinical tool.

Methods

T1-w and T2-w images were obtained for 17 CTN patients and 17 MSTN patients using a 3 T scanner. Template images were obtained from ICBM152. Multiple sclerosis (MS) plaques in the pons were labelled in MSTN patients. For each patient image, a Gaussian curve was fitted to the histogram of its intensity distribution, and transformed to match the Gaussian curve of its template image.

Results

After standardization, the structural contrast of the patient image and its histogram more closely resembled the ICBM152 template. Moreover, there was reduced variability in the histogram peaks of the gray and white matter between patients after standardization (p < 0.001). MM intensities were decreased within MS plaques, compared to normal-appearing white matter (NAWM) in MSTN patients (p < 0.001) and its corresponding regions in CTN patients (p < 0.001).

Conclusions

Images intensities are calibrated according to a mathematic relationship between the intensities of the patient image and its template. Reduced variability among histogram peaks allows for interpretation of tissue-specific intensity and facilitates quantitative analysis. The resultant MMs facilitate comparisons of myelin content between different regions of the brain and between different patients in vivo. MM analysis revealed reduced myelin content in MS plaques compared to its corresponding regions in CTN patients and its surrounding NAWM in MSTN patients. Thus, the standardized MM serves as a non-invasive, easily-automated tool that can be feasibly applied to clinical populations for quantitative analyses of myelin content.

Deep Multi-Scale 3D Convolutional Neural Network (CNN) for MRI Gliomas Brain Tumor Classification

Accurate and fully automatic brain tumor grading from volumetric 3D magnetic resonance imaging (MRI) is an essential procedure in the field of medical imaging analysis for full assistance of neuroradiology during clinical diagnosis. We propose, in this paper, an efficient and fully automatic deep multi-scale three-dimensional convolutional neural network (3D CNN) architecture for glioma brain tumor classification into low-grade gliomas (LGG) and high-grade gliomas (HGG) using the whole volumetric T1-Gado MRI sequence. Based on a 3D convolutional layer and a deep network, via small kernels, the proposed architecture has the potential to merge both the local and global contextual information with reduced weights. To overcome the data heterogeneity, we proposed a preprocessing technique based on intensity normalization and adaptive contrast enhancement of MRI data. Furthermore, for an effective training of such a deep 3D network, we used a data augmentation technique. The paper studied the impact of the proposed preprocessing and data augmentation on classification accuracy.Quantitative evaluations, over the well-known benchmark (Brats-2018), attest that the proposed architecture generates the most discriminative feature map to distinguish between LG and HG gliomas compared with 2D CNN variant. The proposed approach offers promising results outperforming the recently supervised and unsupervised state-of-the-art approaches by achieving an overall accuracy of 96.49% using the validation dataset. The obtained experimental results confirm that adequate MRI's preprocessing and data augmentation could lead to an accurate classification when exploiting CNN-based approaches.

Acoustic simulation for transcranial focused ultrasound using GAN-based synthetic CT

Transcranial focused ultrasound (tFUS) is a promising non-invasive technique for treating neurological and psychiatric disorders. One of the challenges for tFUS is the disruption of wave propagation through the skull. Consequently, despite the risks associated with exposure to ionizing radiation, computed tomography (CT) is required to estimate the acoustic transmission through the skull. This study aims to generate synthetic CT (sCT) from T1-weighted magnetic resonance imaging (MRI) and investigate its applicability to tFUS acoustic simulation. We trained a 3D conditional generative adversarial network (3D-cGAN) with 15 subjects. We then assessed image quality with 15 test subjects: mean absolute error (MAE) = 85.72± 9.50 HU (head) and 280.25±24.02 HU (skull), dice coefficient similarity (DSC) = 0.88±0.02 (skull). In terms of skull density ratio (SDR) and skull thickness (ST), no significant difference was found between sCT and real CT (rCT). When the acoustic simulation results of rCT and sCT were compared, the intracranial peak acoustic pressure ratio was found to be less than 4%, and the distance between focal points less than 1 mm.

Impact of Preprocessing and Harmonization Methods on the Removal of Scanner Effects in Brain MRI Radiomic Features

Simple Summary

As a rapid-development research field, radiomics-based analysis has been applied to many clinical problems. However, the reproducibility of the radiomics studies remain challenging especially when data suffers from scanner effects, a kind of non-biological variations introduced by different image acquiring settings. This study aims to investigate how the image preprocessing methods (N4 bias field correction and image resampling) and the harmonization methods (intensity normalization methods working on images and ComBat method working on radiomic features) help to remove the scanner effects and improve the radiomics reproducibility in brain MRI radiomics.

Abstract

In brain MRI radiomics studies, the non-biological variations introduced by different image acquisition settings, namely scanner effects, affect the reliability and reproducibility of the radiomics results. This paper assesses how the preprocessing methods (including N4 bias field correction and image resampling) and the harmonization methods (either the six intensity normalization methods working on brain MRI images or the ComBat method working on radiomic features) help to remove the scanner effects and improve the radiomic feature reproducibility in brain MRI radiomics. The analyses were based on in vitro datasets (homogeneous and heterogeneous phantom data) and in vivo datasets (brain MRI images collected from healthy volunteers and clinical patients with brain tumors). The results show that the ComBat method is essential and vital to remove scanner effects in brain MRI radiomic studies. Moreover, the intensity normalization methods, while not able to remove scanner effects at the radiomic feature level, still yield more comparable MRI images and improve the robustness of the harmonized features to the choice among ComBat implementations.

SA-LuT-Nets: Learning Sample-Adaptive Intensity Lookup Tables for Brain Tumor Segmentation

In clinics, the information about the appearance and location of brain tumors is essential to assist doctors in diagnosis and treatment. Automatic brain tumor segmentation on the images acquired by magnetic resonance imaging (MRI) is a common way to attain this information. However, MR images are not quantitative and can exhibit significant variation in signal depending on a range of factors, which increases the difficulty of training an automatic segmentation network and applying it to new MR images. To deal with this issue, this paper proposes to learn a sample-adaptive intensity lookup table (LuT) that dynamically transforms the intensity contrast of each input MR image to adapt to the following segmentation task. Specifically, the proposed deep SA-LuT-Net framework consists of a LuT module and a segmentation module, trained in an end-to-end manner: the LuT module learns a sample-specific nonlinear intensity mapping function through communication with the segmentation module, aiming at improving the final segmentation performance. In order to make the LuT learning sample-adaptive, we parameterize the intensity mapping function by exploring two families of non-linear functions (i.e., piece-wise linear and power functions) and predict the function parameters for each given sample. These sample-specific parameters make the intensity mapping adaptive to samples. We develop our SA-LuT-Nets separately based on two backbone networks for segmentation, i.e., DMFNet and the modified 3D Unet, and validate them on BRATS2018 and BRATS2019 datasets for brain tumor segmentation. Our experimental results clearly demonstrate the superior performance of the proposed SA-LuT-Nets using either single or multiple MR modalities. It not only significantly improves the two baselines (DMFNet and the modified 3D Unet), but also wins a set of state-of-the-art segmentation methods. Moreover, we show that, the LuTs learnt using one segmentation model could also be applied to improving the performance of another segmentation model, indicating the general segmentation information captured by LuTs.

Transductive Transfer Learning for Domain Adaptation in Brain Magnetic Resonance Image Segmentation

Segmentation of brain images from Magnetic Resonance Images (MRI) is an indispensable step in clinical practice. Morphological changes of sub-cortical brain structures and quantification of brain lesions are considered biomarkers of neurological and neurodegenerative disorders and used for diagnosis, treatment planning, and monitoring disease progression. In recent years, deep learning methods showed an outstanding performance in medical image segmentation. However, these methods suffer from generalisability problem due to inter-centre and inter-scanner variabilities of the MRI images. The main objective of the study is to develop an automated deep learning segmentation approach that is accurate and robust to the variabilities in scanner and acquisition protocols. In this paper, we propose a transductive transfer learning approach for domain adaptation to reduce the domain-shift effect in brain MRI segmentation. The transductive scenario assumes that there are sets of images from two different domains: (1) source-images with manually annotated labels; and (2) target-images without expert annotations. Then, the network is jointly optimised integrating both source and target images into the transductive training process to segment the regions of interest and to minimise the domain-shift effect. We proposed to use a histogram loss in the feature level to carry out the latter optimisation problem. In order to demonstrate the benefit of the proposed approach, the method has been tested in two different brain MRI image segmentation problems using multi-centre and multi-scanner databases for: (1) sub-cortical brain structure segmentation; and (2) white matter hyperintensities segmentation. The experiments showed that the segmentation performance of a pre-trained model could be significantly improved by up to 10%. For the first segmentation problem it was possible to achieve a maximum improvement from 0.680 to 0.799 in average Dice Similarity Coefficient (DSC) metric and for the second problem the average DSC improved from 0.504 to 0.602. Moreover, the improvements after domain adaptation were on par or showed better performance compared to the commonly used traditional unsupervised segmentation methods (FIRST and LST), also achieving faster execution time. Taking this into account, this work presents one more step toward the practical implementation of deep learning algorithms into the clinical routine.

Gaussian smoothing and modified histogram normalization methods to improve neural-biomarker interpretations for dyslexia classification mechanism

Achieving biologically interpretable neural-biomarkers and features from neuroimaging datasets is a challenging task in an MRI-based dyslexia study. This challenge becomes more pronounced when the needed MRI datasets are collected from multiple heterogeneous sources with inconsistent scanner settings. This study presents a method of improving the biological interpretation of dyslexia's neural-biomarkers from MRI datasets sourced from publicly available open databases. The proposed system utilized a modified histogram normalization (MHN) method to improve dyslexia neural-biomarker interpretations by mapping the pixels' intensities of low-quality input neuroimages to range between the low-intensity region of interest (ROIlow) and high-intensity region of interest (ROIhigh) of the high-quality image. This was achieved after initial image smoothing using the Gaussian filter method with an isotropic kernel of size 4mm. The performance of the proposed smoothing and normalization methods was evaluated based on three image post-processing experiments: ROI segmentation, gray matter (GM) tissues volume estimations, and deep learning (DL) classifications using Computational Anatomy Toolbox (CAT12) and pre-trained models in a MATLAB working environment. The three experiments were preceded by some pre-processing tasks such as image resizing, labelling, patching, and non-rigid registration. Our results showed that the best smoothing was achieved at a scale value, σ = 1.25 with a 0.9% increment in the peak-signal-to-noise ratio (PSNR). Results from the three image post-processing experiments confirmed the efficacy of the proposed methods. Evidence emanating from our analysis showed that using the proposed MHN and Gaussian smoothing methods can improve comparability of image features and neural-biomarkers of dyslexia with a statistically significantly high disc similarity coefficient (DSC) index, low mean square error (MSE), and improved tissue volume estimations. After 10 repeated 10-fold cross-validation, the highest accuracy achieved by DL models is 94.7% at a 95% confidence interval (CI) level. Finally, our finding confirmed that the proposed MHN method significantly outperformed the normalization method of the state-of-the-art histogram matching.

Deep Learning Corpus Callosum Segmentation as a Neurodegenerative Marker in Multiple Sclerosis

BACKGROUND AND PURPOSE

Corpus callosum atrophy is a sensitive biomarker of multiple sclerosis (MS) neurodegeneration but typically requires manual 2D or volumetric 3D-based segmentations. We developed a supervised machine learning algorithm, DeepnCCA, for corpus callosum segmentation and relate callosal morphology to clinical disability using conventional MRI scans collected in clinical routine.

METHODS

In a prospective study of 553 MS patients with 704 acquisitions, 200 unique 2D T₂ -weighted MRI scans were delineated to develop, train, and validate DeepnCCA. Comparative FreeSurfer segmentations were obtained in 504 3D T₁ -weighted scans. Both FreeSurfer and DeepnCCA outputs were correlated with clinical disability. Using principal component analysis of the DeepnCCA output, the morphological changes were explored in relation to clinical disease burden.

RESULTS

DeepnCCA and manual segmentations had high similarity (Dice coefficients 98.1 ± .11%, 89.3 ± .76%, for intracranial and corpus callosum area, respectively through 10-fold cross-validation). DeepnCCA had numerically stronger correlations with cognitive and physical disability as compared to FreeSurfer: Expanded disability status scale (EDSS) ±6 months (r = -.22 P = .002; r = -.17, P = .013), future EDSS (r = -.26, P<.001; r = -.17, P = .012), and future symbol digit modalities test (r = .26, P = .001; r = .24, P = .003). The corpus callosum became thinner with increasing cognitive and physical disability. Increasing physical disability, additionally, significantly correlated with a more angled corpus callosum.

CONCLUSIONS

DeepnCCA (https://github.com/plattenmichael/DeepnCCA/) is an openly available tool that can provide fast and accurate corpus callosum measurements applicable to large MS cohorts, potentially suitable for monitoring disease progression and therapy response.

Radiomics and Dosiomics for Predicting Local Control after Carbon-Ion Radiotherapy in Skull-Base Chordoma

Skull-base chordoma (SBC) can be treated with carbon ion radiotherapy (CIRT) to improve local control (LC). The study aimed to explore the role of multi-parametric radiomic, dosiomic and clinical features as prognostic factors for LC in SBC patients undergoing CIRT. Before CIRT, 57 patients underwent MR and CT imaging, from which tumour contours and dose maps were obtained. MRI and CT-based radiomic, and dosiomic features were selected and fed to two survival models, singularly or by combining them with clinical factors. Adverse LC was given by in-field recurrence or tumour progression. The dataset was split in development and test sets and the models' performance evaluated using the concordance index (C-index). Patients were then assigned a low- or high-risk score. Survival curves were estimated, and risk groups compared through log-rank tests (after Bonferroni correction α = 0.0083). The best performing models were built on features describing tumour shape and dosiomic heterogeneity (median/interquartile range validation C-index: 0.80/024 and 0.79/0.26), followed by combined (0.73/0.30 and 0.75/0.27) and CT-based models (0.77/0.24 and 0.64/0.28). Dosiomic and combined models could consistently stratify patients in two significantly different groups. Dosiomic and multi-parametric radiomic features showed to be promising prognostic factors for LC in SBC treated with CIRT.

Precise enhancement quantification in post-operative MRI as an indicator of residual tumor impact is associated with survival in patients with glioblastoma

Glioblastoma is the most common primary brain tumor. Standard therapy consists of maximum safe resection combined with adjuvant radiochemotherapy followed by chemotherapy with temozolomide, however prognosis is extremely poor. Assessment of the residual tumor after surgery and patient stratification into prognostic groups (i.e., by tumor volume) is currently hindered by the subjective evaluation of residual enhancement in medical images (magnetic resonance imaging [MRI]). Furthermore, objective evidence defining the optimal time to acquire the images is lacking. We analyzed 144 patients with glioblastoma, objectively quantified the enhancing residual tumor through computational image analysis and assessed the correlation with survival. Pathological enhancement thickness on post-surgical MRI correlated with survival (hazard ratio: 1.98, p < 0.001). The prognostic value of several imaging and clinical variables was analyzed individually and combined (radiomics AUC 0.71, p = 0.07; combined AUC 0.72, p < 0.001). Residual enhancement thickness and radiomics complemented clinical data for prognosis stratification in patients with glioblastoma. Significant results were only obtained for scans performed between 24 and 72 h after surgery, raising the possibility of confounding non-tumor enhancement in very early post-surgery MRI. Regarding the extent of resection, and in agreement with recent studies, the association between the measured tumor remnant and survival supports maximal safe resection whenever possible.

Image Contrast, Image Pre-Processing, and T1 Mapping Affect MRI Radiomic Feature Repeatability in Patients with Colorectal Cancer Liver Metastases

Imaging biomarkers require technical, biological, and clinical validation to be translated into robust tools in research or clinical settings. This study contributes to the technical validation of radiomic features from magnetic resonance imaging (MRI) by evaluating the repeatability of features from four MR sequences: pre-contrast T1- and T2-weighted images, pre-contrast quantitative T1 maps (qT1), and contrast-enhanced T1-weighted images. Fifty-one patients with colorectal cancer liver metastases were scanned twice, up to 7 days apart. Repeatability was quantified using the intraclass correlation coefficient (ICC) and repeatability coefficient (RC), and the impact of non-Gaussian feature distributions and image normalisation was evaluated. Most radiomic features had non-Gaussian distributions, but Box-Cox transformations enabled ICCs and RCs to be calculated appropriately for an average of 97% of features across sequences. ICCs ranged from 0.30 to 0.99, with volume and other shape features tending to be most repeatable; volume ICC > 0.98 for all sequences. 19% of features from non-normalised images exhibited significantly different ICCs in pair-wise sequence comparisons. Normalisation tended to increase ICCs for pre-contrast T1- and T2-weighted images, and decrease ICCs for qT1 maps. RCs tended to vary more between sequences than ICCs, showing that evaluations of feature performance depend on the chosen metric. This work suggests that feature-specific repeatability, from specific combinations of MR sequence and pre-processing steps, should be evaluated to select robust radiomic features as biomarkers in specific studies. In addition, as different repeatability metrics can provide different insights into a specific feature, consideration of the appropriate metric should be taken in a study-specific context.

Impact of different scanners and acquisition parameters on robustness of MR radiomics features based on women's cervix

MR Radiomics based on cervical lesions from one single scanner has achieved promising results. However, it is a challenge to achieve clinical translation. Considering multi-scanners and non-uniform scanning parameters from different centers in a real-world medical scenario, we should first identify the influence of such conditions on the robustness of MR radiomics features (RFs) based on the female cervix. In this study, 9 healthy female volunteers were enrolled and 3 kiwis were selected as references. Each of them underwent T2 weighted imaging in three different 3.0-T MR scanners with uniform acquisition parameters, and in one MR scanner with various scanning parameters. A total of 396 RFs were extracted from their images with and without decile intensity normalization. The RFs’ reproducibility was evaluated by coefficient of variation (CV) and quartile coefficient of dispersion (QCD). Representative features were selected using the hierarchical cluster analysis and their discrimination abilities were estimated by ROC analysis through retrospective comparison with the junctional zone and the outer muscular layer of healthy cervix in patients (n = 58) with leiomyoma. This study showed that only a few RFs were robust across different MR scanners and acquisition parameters based on females’ cervix, which might be improved by decile intensity normalization method.

A radiomics pipeline dedicated to Breast MRI: validation on a multi-scanner phantom study

OBJECTIVE

Quantitative analysis in MRI is challenging due to variabilities in intensity distributions across patients, acquisitions and scanners and suffers from bias field inhomogeneity. Radiomic studies are impacted by these effects that affect radiomic feature values. This paper describes a dedicated pipeline to increase reproducibility in breast MRI radiomic studies.

MATERIALS AND METHODS

T1, T2, and T1-DCE MR images of two breast phantoms were acquired using two scanners and three dual breast coils. Images were retrospectively corrected for bias field inhomogeneity and further normalised using Z score or histogram matching. Extracted radiomic features were harmonised between coils by the ComBat method. The whole pipeline was assessed qualitatively and quantitatively using statistical comparisons on two series of radiomic feature values computed in the gel mimicking the normal breast tissue or in dense lesions.

RESULTS

Intra and inter-acquisition variabilities were strongly reduced by the standardisation pipeline. Harmonisation by ComBat lowered the percentage of radiomic features significantly different between the three coils from 87% after bias field correction and MR normalisation to 3% in the gel, while preserving or improving performance of lesion classification in the phantoms.

DISCUSSION

A dedicated standardisation pipeline was developed to reduce variabilities in breast MRI, which paves the way for robust multi-scanner radiomic studies but needs to be assessed on patient data.

Multiple Sclerosis lesions detection by a hybrid Watershed-Clustering algorithm

BACKGROUND

Computer Aided Diagnosis (CAD) systems have been developing in the last years with the aim of helping the diagnosis and monitoring of several diseases. We present a novel CAD system based on a hybrid Watershed-Clustering algorithm for the detection of lesions in Multiple Sclerosis.

METHODS

Magnetic Resonance Imaging scans (FLAIR sequences without gadolinium) of 20 patients affected by Multiple Sclerosis with hyperintense lesions were studied. The CAD system consisted of the following automated processing steps: images recording, automated segmentation based on the Watershed algorithm, detection of lesions, extraction of both dynamic and morphological features, and classification of lesions by Cluster Analysis.

RESULTS

The investigation was performed on 316 suspect regions including 255 lesion and 61 non-lesion cases. The Receiver Operating Characteristic analysis revealed a highly significant difference between lesions and non-lesions; the diagnostic accuracy was 87% (95% CI: 0.83-0.90), with an appropriate cut-off of 192.8; the sensitivity was 77% and the specificity was 87%.

CONCLUSIONS

In conclusion, we developed a CAD system by using a modified algorithm for automated image segmentation which may discriminate MS lesions from non-lesions. The proposed method generates a detection out-put that may be support the clinical evaluation.

Unsupervised MRI Homogenization: Application to Pediatric Anterior Visual Pathway Segmentation

Deep learning strategies have become ubiquitous optimization tools for medical image analysis. With the appropriate amount of data, these approaches outperform classic methodologies in a variety of image processing tasks. However, rare diseases and pediatric imaging often lack extensive data. Specially, MRI are uncommon because they require sedation in young children. Moreover, the lack of standardization in MRI protocols introduces a strong variability between different datasets. In this paper, we present a general deep learning architecture for MRI homogenization that also provides the segmentation map of an anatomical region of interest. Homogenization is achieved using an unsupervised architecture based on variational autoencoder with cycle generative adversarial networks, which learns a common space (i.e. a representation of the optimal imaging protocol) using an unpaired image-to-image translation network. The segmentation is simultaneously generated by a supervised learning strategy. We evaluated our method segmenting the challenging anterior visual pathway using three brain T1-weighted MRI datasets (variable protocols and vendors). Our method significantly outperformed a non-homogenized multi-protocol U-Net.

Automatic Prediction of MGMT Status in Glioblastoma via Deep Learning-Based MR Image Analysis

Methylation of the O⁶-methylguanine methyltransferase (MGMT) gene promoter is correlated with the effectiveness of the current standard of care in glioblastoma patients. In this study, a deep learning pipeline is designed for automatic prediction of MGMT status in 87 glioblastoma patients with contrast-enhanced T1W images and 66 with fluid-attenuated inversion recovery(FLAIR) images. The end-to-end pipeline completes both tumor segmentation and status classification. The better tumor segmentation performance comes from FLAIR images (Dice score, 0.897 ± 0.007) compared to contrast-enhanced T1WI (Dice score, 0.828 ± 0.108), and the better status prediction is also from the FLAIR images (accuracy, 0.827 ± 0.056; recall, 0.852 ± 0.080; precision, 0.821 ± 0.022; and F ₁ score, 0.836 ± 0.072). This proposed pipeline not only saves the time in tumor annotation and avoids interrater variability in glioma segmentation but also achieves good prediction of MGMT methylation status. It would help find molecular biomarkers from routine medical images and further facilitate treatment planning.

Different scaling of linear models and deep learning in UKBiobank brain images versus machine-learning datasets

Recently, deep learning has unlocked unprecedented success in various domains, especially using images, text, and speech. However, deep learning is only beneficial if the data have nonlinear relationships and if they are exploitable at available sample sizes. We systematically profiled the performance of deep, kernel, and linear models as a function of sample size on UKBiobank brain images against established machine learning references. On MNIST and Zalando Fashion, prediction accuracy consistently improves when escalating from linear models to shallow-nonlinear models, and further improves with deep-nonlinear models. In contrast, using structural or functional brain scans, simple linear models perform on par with more complex, highly parameterized models in age/sex prediction across increasing sample sizes. In sum, linear models keep improving as the sample size approaches ~10,000 subjects. Yet, nonlinearities for predicting common phenotypes from typical brain scans remain largely inaccessible to the examined kernel and deep learning methods.

Intensity warping for multisite MRI harmonization

In multisite neuroimaging studies there is often unwanted technical variation across scanners and sites. These "scanner effects" can hinder detection of biological features of interest, produce inconsistent results, and lead to spurious associations. We propose mica (multisite image harmonization by cumulative distribution function alignment), a tool to harmonize images taken on different scanners by identifying and removing within-subject scanner effects. Our goals in the present study were to (1) establish a method that removes scanner effects by leveraging multiple scans collected on the same subject, and, building on this, (2) develop a technique to quantify scanner effects in large multisite studies so these can be reduced as a preprocessing step. We illustrate scanner effects in a brain MRI study in which the same subject was measured twice on seven scanners, and assess our method's performance in a second study in which ten subjects were scanned on two machines. We found that unharmonized images were highly variable across site and scanner type, and our method effectively removed this variability by aligning intensity distributions. We further studied the ability to predict image harmonization results for a scan taken on an existing subject at a new site using cross-validation.

Standardization of brain MR images across machines and protocols: bridging the gap for MRI-based radiomics

Radiomics relies on the extraction of a wide variety of quantitative image-based features to provide decision support. Magnetic resonance imaging (MRI) contributes to the personalization of patient care but suffers from being highly dependent on acquisition and reconstruction parameters. Today, there are no guidelines regarding the optimal pre-processing of MR images in the context of radiomics, which is crucial for the generalization of published image-based signatures. This study aims to assess the impact of three different intensity normalization methods (Nyul, WhiteStripe, Z-Score) typically used in MRI together with two methods for intensity discretization (fixed bin size and fixed bin number). The impact of these methods was evaluated on first- and second-order radiomics features extracted from brain MRI, establishing a unified methodology for future radiomics studies. Two independent MRI datasets were used. The first one (DATASET1) included 20 institutional patients with WHO grade II and III gliomas who underwent post-contrast 3D axial T1-weighted (T1w-gd) and axial T2-weighted fluid attenuation inversion recovery (T2w-flair) sequences on two different MR devices (1.5 T and 3.0 T) with a 1-month delay. Jensen–Shannon divergence was used to compare pairs of intensity histograms before and after normalization. The stability of first-order and second-order features across the two acquisitions was analysed using the concordance correlation coefficient and the intra-class correlation coefficient. The second dataset (DATASET2) was extracted from the public TCIA database and included 108 patients with WHO grade II and III gliomas and 135 patients with WHO grade IV glioblastomas. The impact of normalization and discretization methods was evaluated based on a tumour grade classification task (balanced accuracy measurement) using five well-established machine learning algorithms. Intensity normalization highly improved the robustness of first-order features and the performances of subsequent classification models. For the T1w-gd sequence, the mean balanced accuracy for tumour grade classification was increased from 0.67 (95% CI 0.61–0.73) to 0.82 (95% CI 0.79–0.84, P = .006), 0.79 (95% CI 0.76–0.82, P = .021) and 0.82 (95% CI 0.80–0.85, P = .005), respectively, using the Nyul, WhiteStripe and Z-Score normalization methods compared to no normalization. The relative discretization makes unnecessary the use of intensity normalization for the second-order radiomics features. Even if the bin number for the discretization had a small impact on classification performances, a good compromise was obtained using the 32 bins considering both T1w-gd and T2w-flair sequences. No significant improvements in classification performances were observed using feature selection. A standardized pre-processing pipeline is proposed for the use of radiomics in MRI of brain tumours. For models based on first- and second-order features, we recommend normalizing images with the Z-Score method and adopting an absolute discretization approach. For second-order feature-based signatures, relative discretization can be used without prior normalization. In both cases, 32 bins for discretization are recommended. This study may pave the way for the multicentric development and validation of MR-based radiomics biomarkers.

Radiomic prediction of mutation status based on MR imaging of lung cancer brain metastases

Lung cancer metastases comprise most of all brain metastases in adults and most brain metastases are diagnosed by magnetic resonance (MR) scans. The purpose of this study was to conduct an MR imaging-based radiomic analysis of brain metastatic lesions from patients with primary lung cancer to classify mutational status of the metastatic disease. We retrospectively identified lung cancer patients with brain metastases treated at our institution between 2009 and 2017 who underwent genotype testing of their primary lung cancer. Brain MR Images were used for segmentation of enhancing tumors and peritumoral edema, and for radiomic feature extraction. The most relevant radiomic features were identified and used with clinical data to train random forest classifiers to classify the mutation status. Of 110 patients in the study cohort (mean age 57.51 ± 12.32 years; M: F = 37:73), 75 had an EGFR mutation, 21 had an ALK translocation, and 15 had a KRAS mutation. One patient had both ALK translocation and EGFR mutation. Majority of radiomic features most relevant for mutation classification were textural. Model building using both radiomic features and clinical data yielded more accurate classifications than using either alone. For classification of EGFR, ALK, and KRAS mutation status, the model built with both radiomic features and clinical data resulted in area-under-the-curve (AUC) values based on cross-validation of 0.912, 0.915, and 0.985, respectively. Our study demonstrated that MR imaging-based radiomic analysis of brain metastases in patients with primary lung cancer may be used to classify mutation status. This approach may be useful for devising treatment strategies and informing prognosis.

Effects of MRI image normalization techniques in prostate cancer radiomics

The variance in intensities of MRI scans is a fundamental impediment for quantitative MRI analysis. Intensity values are not only highly dependent on acquisition parameters, but also on the subject and body region being scanned. This warrants the need for image normalization techniques to ensure that intensity values are consistent within tissues across different subjects and visits. Many intensity normalization methods have been developed and proven successful for the analysis of brain pathologies, but evaluation of these methods for images of the prostate region is lagging. In this paper, we compare four different normalization methods on 49 T2-w scans of prostate cancer patients: 1) the well-established histogram normalization, 2) the generalized scale normalization, 3) an extension of generalized scale normalization called generalized ball-scale normalization, and 4) a custom normalization based on healthy prostate tissue intensities. The methods are compared qualitatively and quantitatively in terms of behaviors of intensity distributions as well as impact on radiomic features. Our findings suggest that normalization based on prior knowledge of the healthy prostate tissue intensities may be the most effective way of acquiring the desired properties of normalized images. In addition, the histogram normalization method outperform the generalized scale and generalized ball-scale methods which have proven superior for other body regions.

Automated air region delineation on MRI for synthetic CT creation

Automatically and accurately separating air from other low signal regions (especially bone, liver, etc) in an MRI is difficult because these tissues produce similar MR intensities, resulting in errors in synthetic CT generation for MRI-based radiation therapy planning. This work aims to develop a technique to accurately and automatically determine air-regions for MR-guided adaptive radiation therapy. CT and MRI scans (T2-weighted) of phantoms with fabricated air-cavities and abdominal cancer patients were used to establish an MR intensity threshold for air delineation. From the phantom data, air/tissue boundaries in MRI were identified by CT-MRI registration. A formula relating the MRI intensities of air and surrounding materials was established to auto-threshold air-regions. The air-regions were further refined by using quantitative image texture features. A naive Bayesian classifier was trained using the extracted features with a leave-one-out cross validation technique to differentiate air from non-air voxels. The multi-step air auto-segmentation method was tested against the manually segmented air-regions. The dosimetry impacts of the air-segmentation methods were studied. Air-regions in the abdomen can be segmented on MRI within 1 mm accuracy using a multi-step auto-segmentation method as compared to manually delineated contours. The air delineation based on the MR threshold formula was improved using the MRI texture differences between air and tissues, as judged by the area under the receiver operating characteristic curve of 81% when two texture features (autocorrelation and contrast) were used. The performance increased to 82% with using three features (autocorrelation, sum-variance, and contrast). Dosimetric analysis showed no significant difference between the auto-segmentation and manual MR air delineation on commonly used dose volume parameters. The proposed techniques consisting of intensity-based auto-thresholding and image texture-based voxel classification can automatically and accurately segment air-regions on MRI, allowing synthetic CT to be generated quickly and precisely for MR-guided adaptive radiation therapy.