Quantitative and textural analysis of magnetization transfer and diffusion images in the early detection of brain metastases

Brain metastasis detection using machine learning: a systematic review and meta-analysis

BACKGROUND

Accurate detection of brain metastasis (BM) is important for cancer patients. We aimed to systematically review the performance and quality of machine-learning-based BM detection on MRI in the relevant literature.

METHODS

A systematic literature search was performed for relevant studies reported before April 27, 2020. We assessed the quality of the studies using modified tailored questionnaires of the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) criteria and the Checklist for Artificial Intelligence in Medical Imaging (CLAIM). Pooled detectability was calculated using an inverse-variance weighting model.

RESULTS

A total of 12 studies were included, which showed a clear transition from classical machine learning (cML) to deep learning (DL) after 2018. The studies on DL used a larger sample size than those on cML. The cML and DL groups also differed in the composition of the dataset, and technical details such as data augmentation. The pooled proportions of detectability of BM were 88.7% (95% CI, 84-93%) and 90.1% (95% CI, 84-95%) in the cML and DL groups, respectively. The false-positive rate per person was lower in the DL group than the cML group (10 vs 135, P < 0.001). In the patient selection domain of QUADAS-2, three studies (25%) were designated as high risk due to non-consecutive enrollment and arbitrary exclusion of nodules.

CONCLUSION

A comparable detectability of BM with a low false-positive rate per person was found in the DL group compared with the cML group. Improvements are required in terms of quality and study design.

Radiomics analysis of lung CT image for the early detection of metastases in patients with breast cancer: preliminary findings from a retrospective cohort study

OBJECTIVES

To investigate whether subtle changes in radiomics features are present in lung CT images prior to the development of CT-detectable lung metastases in patients with breast cancer.

METHODS

Thirty-three radiomics features were measured in the metastasis region (MR) and in matched contralateral tissues (non-metastasis region, NMR) of 29 breast cancer patients at the last CT scan, as well as in the corresponding regions of the patients' pre-metastasis scan (pre-MR and pre-NMR). We also compared them with normal lung tissues (control group, CG) from 29 healthy volunteers. Then, 8 patients from the 29 patients with lung metastases and 8 patients who did not develop lung metastases were chosen for further study of the correlation between radiomics parameters and tumor growth.

RESULTS

In the MR vs. NMR and MR vs. CG groups, almost all radiomics features were significantly different. Twenty-six parameters showed significant differences between the pre-MRs and pre-NMRs. Linear fitting demonstrated a significant correlation between 5 features and tumor growth in the metastasis group, but not in the non-metastasis group. Among them, run percentage was the most representative feature. The calculated area under curves (AUCs), based on run percentage for the classification of metastasis and pre-metastasis, were 0.954 and 0.852, respectively.

CONCLUSIONS

Radiomics features may allow early detection of lung metastases before they become visually detectable, and the feature run percentage may be a promising image surrogate marker for the microinvasion of tumor cells into the lung tissue.

KEY POINTS

• The significant differences in radiomics features between pre-MR and pre-NMR are critical for the early detection of lung metastases. • Five radiomics features show a correlation with tumor growth. • The radiomics feature run percentage may be a potential imaging biomarker for the early detection of lung metastases.

Brain metastases from lung cancer with neuropsychiatric symptoms as the first symptoms

Background

To investigate the neuropsychiatric symptoms and their treatment and outcomes in lung cancer patients with brain metastases (BM), with an attempt to achieve early detection and prompt management of these symptoms.

Methods

Ten lung cancer patients (8 males and 2 females) with BMs who were treated in our center from 2013 to 2019 were enrolled in this analysis. Without exception, all 10 patients presented with chief complaints of neuropsychiatric symptoms, and BMs were eventually diagnosed. Appropriate treatments were offered, and all patients were followed up.

Results

Two patients died (case 5 died of sudden massive hemoptysis, and case 6 died after his families refused to receive the invasive treatment). Data on 3- and 5-year survival have been obtained from one patient each. The average follow-up duration was 19.4 months (except that two patients were hospitalized only once, and one patient received the second follow-up visit only 9 days after the first visit).

Conclusions

The possibility of BM from lung cancer should be considered when a lung cancer patient develops neuropsychiatric symptoms, and timely diagnosis treatment should be arranged accordingly.

Discrimination Between Solitary Brain Metastasis and Glioblastoma Multiforme by Using ADC-Based Texture Analysis: A Comparison of Two Different ROI Placements

RATIONALE AND OBJECTIVES

To explore the value of texture analysis based on the apparent diffusion coefficient (ADC) value and the effect of region of interest (ROI) placements in distinguishing glioblastoma multiforme (GBM) from solitary brain metastasis (sMET).

MATERIALS AND METHODS

Sixty-two patients with pathologically confirmed GBM (n = 36) and sMET (n = 26) were retrospectively included. All patients underwent diffusion-weighted imaging with b values of 0 and 1000 s/mm², and the ADC maps were generated automatically. ROIs were placed on the largest whole single-slice tumor (ROI1) and the enhanced solid portion (ROI2) of the ADC maps, respectively. The texture feature metrics of the histogram and gray-level co-occurrence matrix were then extracted by using in-house software. The parameters of the texture analysis were compared between GBM and sMET, using the Mann-Whitney U test. A receiver operating characteristic (ROC) curve analysis was performed to determine the best parameters for distinguishing between GBM from sMET.

RESULTS

Homogeneity and the inverse difference moment (IDM) of GBM were significantly higher than those of sMET in both ROIs (ROI1, p = 0.014 for homogeneity and p = 0.048 for IDM; ROI2, p< 0.001 for homogeneity and p = 0.029 for IDM). According to the ROC curve analysis, the area under the ROC curve (AUC) of homogeneity in ROI1 (AUC, 0.682, sensitivity, 72.2%, specificity, 61.5%) was significantly lower than that of ROI2 (AUC, 0.886, sensitivity, 83.3%, specificity, 76.9%; p= 0.012), whereas the IDM showed no statistical significance between two ROIs (p> 0.05).

CONCLUSION

The ADC-based texture analysis can help differentiate GBM from sMET, and the ROI on the solid portion would be recommended to calculate the ADC-based texture metrics.

Advanced Magnetic Resonance Imaging Techniques in Management of Brain Metastases

Brain metastases are the most common intracranial tumors and occur in 20–40% of all cancer patients. Lung cancer, breast cancer, and melanoma are the most frequent primary cancers to develop brain metastases. Treatment options include surgical resection, whole brain radiotherapy, stereotactic radiosurgery, and systemic treatment such as targeted or immune therapy. Anatomical magnetic resonance imaging (MRI) of the tumor (in particular post-Gadolinium T₁-weighted and T₂-weighted FLAIR) provide information about lesion morphology and structure, and are routinely used in clinical practice for both detection and treatment response evaluation for brain metastases. Advanced MRI biomarkers that characterize the cellular, biophysical, micro-structural and metabolic features of tumors have the potential to improve the management of brain metastases from early detection and diagnosis, to evaluating treatment response. Magnetic resonance spectroscopy (MRS), chemical exchange saturation transfer (CEST), quantitative magnetization transfer (qMT), diffusion-based tissue microstructure imaging, trans-membrane water exchange mapping, and magnetic susceptibility weighted imaging (SWI) are advanced MRI techniques that will be reviewed in this article as they pertain to brain metastases.

Innovative Therapeutic Strategies for Effective Treatment of Brain Metastases

Brain metastases are the most prevalent of intracranial malignancies. They are associated with a very poor prognosis and near 100% mortality. This has been the case for decades, largely because we lack effective therapeutics to augment surgery and radiotherapy. Notwithstanding improvements in the precision and efficacy of these life-prolonging treatments, with no reliable options for adjunct systemic therapy, brain recurrences are virtually inevitable. The factors limiting intracranial efficacy of existing agents are both physiological and molecular in nature. For example, heterogeneous permeability, abnormal perfusion and high interstitial pressure oppose the conventional convective delivery of circulating drugs, thus new delivery strategies are needed to achieve uniform drug uptake at therapeutic concentrations. Brain metastases are also highly adapted to their microenvironment, with complex cross-talk between the tumor, the stroma and the neural compartments driving speciation and drug resistance. New strategies must account for resistance mechanisms that are frequently engaged in this milieu, such as HER3 and other receptor tyrosine kinases that become induced and activated in the brain microenvironment. Here, we discuss molecular and physiological factors that contribute to the recalcitrance of these tumors, and review emerging therapeutic strategies, including agents targeting the PI3K axis, immunotherapies, nanomedicines and MRI-guided focused ultrasound for externally controlling drug delivery.

MRI based texture analysis to classify low grade gliomas into astrocytoma and 1p/19q codeleted oligodendroglioma

PURPOSE

Texture analysis performed on MR images can detect quantitative features that are imperceptible to human visual assessment. The purpose of this study was to evaluate the feasibility of texture analysis on preoperative conventional MRI to discriminate between histological subtypes in low-grade gliomas (LGGs), and to determine the utility of texture analysis compared to histogram analysis alone.

METHODS

A total of 41 patients with LGG, 21 astrocytoma and 20 1p/19q codeleted oligodendroglioma were included in this study. Patients were randomly divided into training (60%) and testing (40%) sets. Texture analysis was performed on conventional MRI sequences to obtain the most discriminant factor (MDF) values for both the training and testing data. Receiver operating characteristic (ROC) curve analyses were then performed using the MDF values and 9 histogram parameters in the training data to obtain cut-off values for determining the correct rate of discriminating between astrocytoma and oligodendroglioma in the testing data.

RESULTS

The ROC analyses using MDF values resulted in an area under the curve (AUC) of 0.91 (sensitivity 86%, specificity 87%) for T2w FLAIR, 0.94 (87%, 89%) for ADC, 0.98 (93%, 95%) for T1w, and 0.88 (78%, 86%) for T1w + Gd sequences. Using the best cut-off values, MDF correctly discriminated between the two groups in 94%, 82%, 100%, and 88% of cases in the testing data, respectively. The MDF outperformed all 9 of the histogram parameters.

CONCLUSION

Texture analysis performed on conventional preoperative MRI images can accurately predict histological subtype of LGGs, which would have an impact on clinical management.

Predicting brain metastases for non-small cell lung cancer based on magnetic resonance imaging

In this study the relationship between brain structure and brain metastases (BM) occurrence was analyzed. A model for predicting the time of BM onset in patients with non-small cell lung cancer (NSCLC) was proposed. Twenty patients were used to develop the model, whereas the remaining 69 were used for independent validation and verification of the model. Magnetic resonance images were segmented into cerebrospinal fluid, gray matter (GM), and white matter using voxel-based morphometry. Automatic anatomic labeling template was used to extract 116 brain regions from the GM volume. The elapsed time between the MRI acquisitions and BM diagnosed was analyzed using the least absolute shrinkage and selection operator method. The model was validated using the leave-one-out cross validation (LOOCV) and permutation test. The GM volume of the extracted 11 regions of interest increased with the progression of BM from NSCLC. LOOCV test on the model indicated that the measured and predicted BM onset were highly correlated (r = 0.834, P = 0.0000). For the 69 independent validating patients, accuracy, sensitivity, and specificity of the model for predicting BM occurrence were 70, 75, and 66%, respectively, in 6 months and 74, 82, and 60%, respectively, in 1 year. The extracted brain GM volumes and interval times for BM occurrence were correlated. The established model based on MRI data may reliably predict BM in 6 months or 1 year. Further studies with larger sample size are needed to validate the findings in a clinical setting.