DR. MORAN ARTZI (PhD)
Dr. Artzi's research is focused on quantitative tissue characterization and classification based on advanced methods for MR acquisition and analysis. Studies are conducted on healthy subjects and patients with pathologies of the central nervous system. The aim of her research is to improve patient assessment, therapy response monitoring and prediction of clinical outcomes.
DR. Artzi is an associate investigator of the #Advanced Brain Imaging research team at the TLV-CBF.
Brain Tumor Classification
Lesion classification in patients with high grade brain tumors
Towards quantitative analysis
Radiomics
Longitudinal follow-up of low-grade gliomas
Towards quantitative analysis
Detecting "true" changes in MR parameters in the longitudinal assessment of patients with brain lesion is clinically important and extremely challenging - and has gained much interest in recent years.
In our studies we develop methods to provide reliable and reproducible quantitative measures of MR parameters (Link to Method page). In a recent study we used histogram comparison via Earth Mover's Distance (EMD) analysis to assess longitudinal changes in plasma volume and tissue permeability (two pharmacokinetic parameters extracted from dynamic contrast enhanced (DCE) -MRI).
Preliminary results suggest that the use of this method aids in the identification of significant changes in quantitative MRI, thus providing additional information beyond lesion volume.
Figure2: Longitudinal assessment of a 38 years patient with anaplastic astrocytoma grade 3 (total of 8 scans) (a). Matrices of EMD results obtained for the vp parameter from the NAWM+NAGM (b) and from lesion area (c). vp histograms from the NAWM+NAGM (d) and from lesion area (e).
Radiomics
Radiomics deals with the extraction and analysis of large amounts of quantitative imaging features, which were previously not considered clinically useful. Such analysis provides signatures that can be associated with underlying gene-expression patterns and may improve lesion assessment and prognosis.
This concept is currently being applied in several projects in our lab, investigating patients with low and high grade brain tumors (work in progress).
Figure 3: Illustration of various quantitative features extracted from patients with low grade gliomas.
Longitudinal follow-up of low-grade gliomas
Tracking the progression of low grade gliomas (LGGs) is a challenging task due to their slow growth rate and the involvement of internal tumor components, such as enhancing and cyst regions.
In our studies we develop tools for semi-automatic segmentation and internal classification of optic-pathway gliomas in MRI. The lesion area is classified into enhancing, non-enhancing and cystic components. In addition, the method enables reliable and repeatable tracking of internal tumor components in longitudinal scans, thus improving LGG follow-up.
These studies are performed in collaboration with #Prof. Leo Joskowicz from the Hebrew University.
Key Publications
Automatic Segmentation, Internal Classification, and Follow-up of Optic Pathway Gliomas in MRI.
Semi-automatic segmentation and follow-up of multi-component low-grade tumors in longitudinal MRI studies.
The effect of chemotherapy on optic pathway gliomas and their sub-components: A volumetric MR analysis study.
MRI is the established method for assessment of patients with brain tumors. However, the response assessment in neuro-oncology (RANO) criteria currently rely on conventional imaging, which provides only a rough estimation of enhancing and non-enhancing lesion volumes. These parameters fail to reliably distinguish between different tissue components such as tumor progression versus therapy response, and treatment ramifications are influenced by the correct understanding of the nature of the lesion.
Our studies aim to improve therapy response assessment in patients with brain tumors by providing quantitative and reliable tools using, advanced methods of MR data acquisition, state of the art image processing and machine learning algorithms for data analysis.
Lesion classification in patients with high grade brain tumors
Automatic methods that take advantage of state of the art image processing and machine learning algorithms are used for lesion classification.
In our studies, we classify the lesion area in patients with high grade brain tumors into 4 components: enhancing, non-enhancing and tumor, non-tumor components based on conventional imaging and dynamic-contrast-enhanced (DCE) pharmacokinetic parameters.
This method provides high sensitivity and specificity, and automatically provides volumetric information regarding sub-components of the lesion at the patient-level. In 16% of patients, classification results preceded the radiological diagnosis of patient progression, yielding additional information to that of the standard radiological (RANO) criteria and providing early predictionsof treatment response.
Figure1: Data obtained from a 54 year old patient with anaplastic astrocytoma grade 3. Lesion area was classified into 4 components: enhancing tumor; enhancing non-tumor; non-enhancing tumor; non-enhancing non-tumor.
Key Publications
FLAIR Lesion Segmentation: Application in Patients with Brain Tumors and Acute Ischemic Stroke.
Automatic multi-modal MR tissue classification for the assessment of response to bevacizumab in patients with glioblastoma.
Differentiation between vasogenic-edema versus tumor-infiltrative area in patients with glioblastoma during bevacizumab therapy: a longitudinal MRI study.
Classification of tumor area using combined DCE and DSC MRI in patients with glioblastoma.
Differentiation between treatment-related changesand progressive disease in patients with high grade brain tumorsusing support vector machine classification based on DCE MRI.