Seung-Hak Lee

Differences in early and late mild cognitive impairment tractography using a diffusion tensor MRI.

Diffusion tensor MRI tractography is an imaging tool that can provide information of in-vivo neuronal fiber tracts to assess progress for Alzheimer’s disease (AD). In an effort to detect early AD progression, we focused on distinguishing subgroups within mild cognitive impairment (MCI): early MCI and late MCI. Tractography was applied not only to white matter regions but also neighboring gray matter regions known to be affected by AD. Nerve fibers touching the hippocampus, thalamus, and amygdala in both hemispheres were extracted to quantify limbic system fiber connectivity. Two fiber extraction algorithms, fiber assignment by continuous tracking and the Runge Kutta approach, were applied to an AD imaging database. We computed the number of fibers touching regions of interest as the imaging feature. The imaging feature could distinguish between the MCI subgroups. It was also significantly correlated with a known genetic marker for AD, the apolipoprotein E epsilon 4 allele. The number of fibers might be a useful imaging biomarker to complement conventional region of interest-based biomarkers for AD research.

Functional Connectivity of Child and Adolescent Attention Deficit Hyperactivity Disorder Patients: Correlation with IQ

Attention deficit hyperactivity disorder (ADHD) is a pervasive neuropsychological disorder that affects both children and adolescents. Child and adolescent ADHD patients exhibit different behavioral symptoms such as hyperactivity and impulsivity, but not much connectivity research exists to help explain these differences. We analyzed openly accessible resting-state functional magnetic resonance imaging (rs-fMRI) data on 112 patients (28 child ADHD, 28 adolescent ADHD, 28 child normal control (NC), and 28 adolescent NC). We used group independent component analysis (ICA) and weighted degree values to identify interaction effects of age (child and adolescent) and symptom (ADHD and NC) in brain networks. The frontoparietal network showed significant interaction effects (p = 0.0068). The frontoparietal network is known to be related to hyperactive and impulsive behaviors. Intelligence quotient (IQ) is an important factor in ADHD, and we predicted IQ scores using the results of our connectivity analysis. IQ was predicted using degree centrality values of networks with significant interaction effects of age and symptom. Actual and predicted IQ scores demonstrated significant correlation values, with an error of about 10%. Our study might provide imaging biomarkers for future ADHD and intelligence studies.

Planning for selective amygdalohippocampectomy involving less neuronal fiber damage based on brain connectivity using tractography.

Temporal lobe resection is an important treatment option for epilepsy that involves removal of potentially essential brain regions. Selective amygdalohippocampectomy is a widely performed temporal lobe surgery. We suggest starting the incision for selective amygdalohippocampectomy at the inferior temporal gyrus based on diffusion magnetic resonance imaging (MRI) tractography. Diffusion MRI data from 20 normal participants were obtained from Parkinson′s Progression Markers Initiative (PPMI) database (www.ppmi-info.org). A tractography algorithm was applied to extract neuronal fiber information for the temporal lobe, hippocampus, and amygdala. Fiber information was analyzed in terms of the number of fibers and betweenness centrality. Distances between starting incisions and surgical target regions were also considered to explore the length of the surgical path. Middle temporal and superior temporal gyrus regions have higher connectivity values than the inferior temporal gyrus and thus are not good candidates for starting the incision. The distances between inferior temporal gyrus and surgical target regions were shorter than those between middle temporal gyrus and target regions. Thus, the inferior temporal gyrus is a good candidate for starting the incision. Starting the incision from the inferior temporal gyrus would spare the important (in terms of betweenness centrality values) middle region and shorten the distance to the target regions of the hippocampus and amygdala.

DEWS (DEep White matter hyperintensity Segmentation framework): A fully automated pipeline for detecting small deep white matter hyperintensities in migraineurs

Migraineurs show an increased load of white matter hyperintensities (WMHs) and more rapid deep WMH progression. Previous methods for WMH segmentation have limited efficacy to detect small deep WMHs. We developed a new fully automated detection pipeline, DEWS (DEep White matter hyperintensity Segmentation framework), for small and superficially-located deep WMHs. A total of 148 non-elderly subjects with migraine were included in this study. The pipeline consists of three components: 1) white matter (WM) extraction, 2) WMH detection, and 3) false positive reduction. In WM extraction, we adjusted the WM mask to re-assign misclassified WMHs back to WM using many sequential low-level image processing steps. In WMH detection, the potential WMH clusters were detected using an intensity based threshold and region growing approach. For false positive reduction, the detected WMH clusters were classified into final WMHs and non-WMHs using the random forest (RF) classifier. Size, texture, and multi-scale deep features were used to train the RF classifier. DEWS successfully detected small deep WMHs with a high positive predictive value (PPV) of 0.98 and true positive rate (TPR) of 0.70 in the training and test sets. Similar performance of PPV (0.96) and TPR (0.68) was attained in the validation set. DEWS showed a superior performance in comparison with other methods. Our proposed pipeline is freely available online to help the research community in quantifying deep WMHs in non-elderly adults.

Imaging genetics approach to Parkinson’s disease and its correlation with clinical score

Parkinson’s disease (PD) is a progressive neurodegenerative disorder associated with both underlying genetic factors and neuroimaging findings. Existing neuroimaging studies related to the genome in PD have mostly focused on certain candidate genes. The aim of our study was to construct a linear regression model using both genetic and neuroimaging features to better predict clinical scores compared to conventional approaches. We obtained neuroimaging and DNA genotyping data from a research database. Connectivity analysis was applied to identify neuroimaging features that could differentiate between healthy control (HC) and PD groups. A joint analysis of genetic and imaging information known as imaging genetics was applied to investigate genetic variants. We then compared the utility of combining different genetic variants and neuroimaging features for predicting the Movement Disorder Society-sponsored unified Parkinson’s disease rating scale (MDS-UPDRS) in a regression framework. The associative cortex, motor cortex, thalamus, and pallidum showed significantly different connectivity between the HC and PD groups. Imaging genetics analysis identified PARK2, PARK7, HtrA2, GIGYRF2, and SNCA as genetic variants that are significantly associated with imaging phenotypes. A linear regression model combining genetic and neuroimaging features predicted the MDS-UPDRS with lower error and higher correlation with the actual MDS-UPDRS compared to other models using only genetic or neuroimaging information alone.

Deciphering Clinicoradiologic Phenotype for Thymidylate Synthase Expression Status in Patients with Advanced Lung Adenocarcinoma Using a Radiomics Approach

We aimed to identify predictive clinicoradiologic characteristics of thymidylate synthase (TS) expression status in advanced non-squamous non-small cell lung cancer patients. We reviewed clinicoradiologic features of 169 patients stratified into TS-negative (n = 84) and TS-positive (n = 85) groups, including quantitative CT radiomic features of both primary lung and metastatic lesions from initial CT and PET. Clinical factors including age and smoking history were significantly associated with TS as well as radiomic features. The predictive performance for dichotomizing TS expression status was slightly higher when imaging features of primary lung lesions were added compared to the model based solely on the clinical features, but without statistical significance (10-fold cross-validated AUC = 0.619 and 0.581, respectively; P = 0.425). The predictive performance of clinicoradiologic parameters slightly increased with primary lung lesions only compared to the inclusion of metastatic lesions, but without statistical significance (10-fold cross-validated AUC = 0.619 and 0.554, respectively; P = 0.203). Overall survival was prolonged in the TS-negative group compared to the TS-positive group (P = 0.001). TS-negativity is a potential prognostic biomarker, and our study presents that although CT radiomic features have potential for predicting TS expression status, clinical significance is uncertain. The addition of radiomic features to clinical factors did not show significant improvement in predicting TS-negativity.

Comprehensive Computed Tomography Radiomics Analysis of Lung Adenocarcinoma for Prognostication

BACKGROUND In this era of personalized medicine, there is an expanded demand for advanced imaging biomarkers that reflect the biology of the whole tumor. Therefore, we investigated a large number of computed tomography-derived radiomics features along with demographics and pathology-related variables in patients with lung adenocarcinoma, correlating them with overall survival. MATERIALS AND METHODS Three hundred thirty-nine patients who underwent operation for lung adenocarcinoma were included. Analysis was performed using 161 radiomics features, demographic, and pathologic variables and correlated each with patient survival. Prognostic performance for survival was compared among three models: (a) using only clinicopathological data; (b) using only selected radiomics features; and (c) using both clinicopathological data and selected radiomics features. RESULTS At multivariate analysis, age, pN, tumor size, type of operation, histologic grade, maximum value of the outer 1/3 of the tumor, and size zone variance were statistically significant variables. In particular, maximum value of outer 1/3 of the tumor reflected tumor microenvironment, and size zone variance represented intratumor heterogeneity. Integration of 31 selected radiomics features with clinicopathological variables led to better discrimination performance. CONCLUSION Radiomics approach in lung adenocarcinoma enables utilization of the full potential of medical imaging and has potential to improve prognosis assessment in clinical oncology. IMPLICATIONS FOR PRACTICE Two radiomics features were prognostic for lung cancer survival at multivariate analysis: (a) maximum value of the outer one third of the tumor reflects the tumor microenvironment and (b) size zone variance represents the intratumor heterogeneity. Therefore, a radiomics approach in lung adenocarcinoma enables utilization of the full potential of medical imaging and could play a larger role in clinical oncology.

Clustering approach to identify intratumour heterogeneity combining FDG PET and diffusion-weighted MRI in lung adenocarcinoma

ObjectivesMalignant tumours consist of biologically heterogeneous components; identifying and stratifying those various subregions is an important research topic. We aimed to show the effectiveness of an intratumour partitioning method using clustering to identify highly aggressive tumour subregions, determining prognosis based on pre-treatment PET and DWI in stage IV lung adenocarcinoma.MethodsEighteen patients who underwent both baseline PET and DWI were recruited. Pre-treatment imaging of SUV and ADC values were used to form intensity vectors within manually specified ROIs. We applied k-means clustering to intensity vectors to yield distinct subregions, then chose the subregion that best matched the criteria for high SUV and low ADC to identify tumour subregions with high aggressiveness. We stratified patients into high- and low-risk groups based on subregion volume with high aggressiveness and conducted survival analyses. This approach is referred to as the partitioning approach. For comparison, we computed tumour subregions with high aggressiveness without clustering and repeated the described procedure; this is referred to as the voxel-wise approach.ResultsThe partitioning approach led to high-risk (median SUVmax = 14.25 and median ADC = 1.26x10-3 mm2/s) and low-risk (median SUVmax = 14.64 and median ADC = 1.09x10-3 mm2/s) subgroups. Our partitioning approach identified significant differences in survival between high- and low-risk subgroups (hazard ratio, 4.062, 95% confidence interval, 1.21 – 13.58, p-value: 0.035). The voxel-wise approach did not identify significant differences in survival between high- and low-risk subgroups (p-value: 0.325).ConclusionOur partitioning approach identified intratumour subregions that were predictors of survival.Key Points• Multimodal imaging of PET and DWI is useful for assessing intratumour heterogeneity.• Data-driven clustering identified subregions which might be highly aggressive for lung adenocarcinoma.• The data-driven partitioning results might be predictors of survival.