All Publications


  • Radiomics Can Distinguish Pediatric Supratentorial Embryonal Tumors, High-Grade Gliomas, and Ependymomas. AJNR. American journal of neuroradiology Zhang, M., Tam, L., Wright, J., Mohammadzadeh, M., Han, M., Chen, E., Wagner, M., Nemalka, J., Lai, H., Eghbal, A., Ho, C. Y., Lober, R. M., Cheshier, S. H., Vitanza, N. A., Grant, G. A., Prolo, L. M., Yeom, K. W., Jaju, A. 2022; 43 (4): 603-610

    Abstract

    Pediatric supratentorial tumors such as embryonal tumors, high-grade gliomas, and ependymomas are difficult to distinguish by histopathology and imaging because of overlapping features. We applied machine learning to uncover MR imaging-based radiomics phenotypes that can differentiate these tumor types.Our retrospective cohort of 231 patients from 7 participating institutions had 50 embryonal tumors, 127 high-grade gliomas, and 54 ependymomas. For each tumor volume, we extracted 900 Image Biomarker Standardization Initiative-based PyRadiomics features from T2-weighted and gadolinium-enhanced T1-weighted images. A reduced feature set was obtained by sparse regression analysis and was used as input for 6 candidate classifier models. Training and test sets were randomly allocated from the total cohort in a 75:25 ratio.The final classifier model for embryonal tumor-versus-high-grade gliomas identified 23 features with an area under the curve of 0.98; the sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were 0.85, 0.91, 0.79, 0.94, and 0.89, respectively. The classifier for embryonal tumor-versus-ependymomas identified 4 features with an area under the curve of 0.82; the sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were 0.93, 0.69, 0.76, 0.90, and 0.81, respectively. The classifier for high-grade gliomas-versus-ependymomas identified 35 features with an area under the curve of 0.96; the sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were 0.82, 0.94, 0.82, 0.94, and 0.91, respectively.In this multi-institutional study, we identified distinct radiomic phenotypes that distinguish pediatric supratentorial tumors, high-grade gliomas, and ependymomas with high accuracy. Incorporation of this technique in diagnostic algorithms can improve diagnosis, risk stratification, and treatment planning.

    View details for DOI 10.3174/ajnr.A7481

    View details for PubMedID 35361575

    View details for PubMedCentralID PMC8993189

  • MRI-based radiomics for prognosis of pediatric diffuse intrinsic pontine glioma: an international study. Neuro-oncology advances Tam, L. T., Yeom, K. W., Wright, J. N., Jaju, A., Radmanesh, A., Han, M., Toescu, S., Maleki, M., Chen, E., Campion, A., Lai, H. A., Eghbal, A. A., Oztekin, O., Mankad, K., Hargrave, D., Jacques, T. S., Goetti, R., Lober, R. M., Cheshier, S. H., Napel, S., Said, M., Aquilina, K., Ho, C. Y., Monje, M., Vitanza, N. A., Mattonen, S. A. 2021; 3 (1): vdab042

    Abstract

    Background: Diffuse intrinsic pontine gliomas (DIPGs) are lethal pediatric brain tumors. Presently, MRI is the mainstay of disease diagnosis and surveillance. We identify clinically significant computational features from MRI and create a prognostic machine learning model.Methods: We isolated tumor volumes of T1-post-contrast (T1) and T2-weighted (T2) MRIs from 177 treatment-naive DIPG patients from an international cohort for model training and testing. The Quantitative Image Feature Pipeline and PyRadiomics was used for feature extraction. Ten-fold cross-validation of least absolute shrinkage and selection operator Cox regression selected optimal features to predict overall survival in the training dataset and tested in the independent testing dataset. We analyzed model performance using clinical variables (age at diagnosis and sex) only, radiomics only, and radiomics plus clinical variables.Results: All selected features were intensity and texture-based on the wavelet-filtered images (3 T1 gray-level co-occurrence matrix (GLCM) texture features, T2 GLCM texture feature, and T2 first-order mean). This multivariable Cox model demonstrated a concordance of 0.68 (95% CI: 0.61-0.74) in the training dataset, significantly outperforming the clinical-only model (C = 0.57 [95% CI: 0.49-0.64]). Adding clinical features to radiomics slightly improved performance (C = 0.70 [95% CI: 0.64-0.77]). The combined radiomics and clinical model was validated in the independent testing dataset (C = 0.59 [95% CI: 0.51-0.67], Noether's test P = .02).Conclusions: In this international study, we demonstrate the use of radiomic signatures to create a machine learning model for DIPG prognostication. Standardized, quantitative approaches that objectively measure DIPG changes, including computational MRI evaluation, could offer new approaches to assessing tumor phenotype and serve a future role for optimizing clinical trial eligibility and tumor surveillance.

    View details for DOI 10.1093/noajnl/vdab042

    View details for PubMedID 33977272

  • Effects of Age on White Matter Microstructure in Children With Neurofibromatosis Type 1. Journal of child neurology Tam, L. T., Ng, N. N., McKenna, E. S., Bruckert, L., Yeom, K. W., Campen, C. J. 2021: 8830738211008736

    Abstract

    Children with neurofibromatosis type 1 (NF1) often report cognitive challenges, though the etiology of such remains an area of active investigation. With the advent of treatments that may affect white matter microstructure, understanding the effects of age on white matter aberrancies in NF1 becomes crucial in determining the timing of such therapeutic interventions. A cross-sectional study was performed with diffusion tensor imaging from 18 NF1 children and 26 age-matched controls. Fractional anisotropy was determined by region of interest analyses for both groups over the corpus callosum, cingulate, and bilateral frontal and temporal white matter regions. Two-way analyses of variance were done with both ages combined and age-stratified into early childhood, middle childhood, and adolescence. Significant differences in fractional anisotropy between NF1 and controls were seen in the corpus callosum and frontal white matter regions when ages were combined. When stratified by age, we found that this difference was largely driven by the early childhood (1-5.9 years) and middle childhood (6-11.9 years) age groups, whereas no significant differences were appreciable in the adolescence age group (12-18 years). This study demonstrates age-related effects on white matter microstructure disorganization in NF1, suggesting that the appropriate timing of therapeutic intervention may be in early childhood.

    View details for DOI 10.1177/08830738211008736

    View details for PubMedID 34048307

  • Targeting neuronal activity-regulated neuroligin-3 dependency in high-grade glioma. Nature Venkatesh, H. S., Tam, L. T., Woo, P. J., Lennon, J. n., Nagaraja, S. n., Gillespie, S. M., Ni, J. n., Duveau, D. Y., Morris, P. J., Zhao, J. J., Thomas, C. J., Monje, M. n. 2017; 549 (7673): 533–37

    Abstract

    High-grade gliomas (HGG) are a devastating group of cancers, and represent the leading cause of brain tumour-related death in both children and adults. Therapies aimed at mechanisms intrinsic to glioma cells have translated to only limited success; effective therapeutic strategies will need also to target elements of the tumour microenvironment that promote glioma progression. Neuronal activity promotes the growth of a range of molecularly and clinically distinct HGG types, including adult and paediatric glioblastoma (GBM), anaplastic oligodendroglioma, and diffuse intrinsic pontine glioma (DIPG). An important mechanism that mediates this neural regulation of brain cancer is activity-dependent cleavage and secretion of the synaptic adhesion molecule neuroligin-3 (NLGN3), which promotes glioma proliferation through the PI3K-mTOR pathway. However, the necessity of NLGN3 for glioma growth, the proteolytic mechanism of NLGN3 secretion, and the further molecular consequences of NLGN3 secretion in glioma cells remain unknown. Here we show that HGG growth depends on microenvironmental NLGN3, identify signalling cascades downstream of NLGN3 binding in glioma, and determine a therapeutically targetable mechanism of secretion. Patient-derived orthotopic xenografts of paediatric GBM, DIPG and adult GBM fail to grow in Nlgn3 knockout mice. NLGN3 stimulates several oncogenic pathways, such as early focal adhesion kinase activation upstream of PI3K-mTOR, and induces transcriptional changes that include upregulation of several synapse-related genes in glioma cells. NLGN3 is cleaved from both neurons and oligodendrocyte precursor cells via the ADAM10 sheddase. ADAM10 inhibitors prevent the release of NLGN3 into the tumour microenvironment and robustly block HGG xenograft growth. This work defines a promising strategy for targeting NLGN3 secretion, which could prove transformative for HGG therapy.

    View details for PubMedID 28959975

  • Barriers to Telemedicine Video Visits for Older Adults in Independent Living Facilities: Mixed Methods Cross-sectional Needs Assessment. JMIR aging Mao, A., Tam, L., Xu, A., Osborn, K., Sheffrin, M., Gould, C., Schillinger, E., Martin, M., Mesias, M. 2022; 5 (2): e34326

    Abstract

    Despite the increasing availability of telemedicine video visits during the COVID-19 pandemic, older adults have greater challenges in getting care through telemedicine.We aim to better understand the barriers to telemedicine in community-dwelling older adults to improve the access to and experience of virtual visits.We conducted a mixed methods needs assessment of older adults at two independent living facilities (sites A and B) in Northern California between September 2020 and March 2021. Voluntary surveys were distributed. Semistructured interviews were then conducted with participants who provided contact information. Surveys ascertained participants' preferred devices as well as comfort level, support, and top barriers regarding telephonic and video visits. Qualitative analysis of transcribed interviews identified key themes.Survey respondents' (N=249) average age was 84.6 (SD 6.6) years, and 76.7% (n=191) of the participants were female. At site A, 88.9% (111/125) had a bachelor's degree or beyond, and 99.2% (124/125) listed English as their preferred language. At site B, 42.9% (51/119) had a bachelor's degree or beyond, and 13.4% (16/119) preferred English, while 73.1% (87/119) preferred Mandarin. Regarding video visits, 36.5% (91/249) of all participants felt comfortable connecting with their health care team through video visits. Regarding top barriers, participants at site A reported not knowing how to connect to the platform (30/125, 24%), not being familiar with the technology (28/125, 22.4%), and having difficulty hearing (19/125, 15.2%), whereas for site B, the top barriers were not being able to speak English well (65/119, 54.6%), lack of familiarity with technology and the internet (44/119, 36.9%), and lack of interest in seeing providers outside of the clinic (42/119, 35.3%). Three key themes emerged from the follow-up interviews (n=15): (1) the perceived limitations of video visits, (2) the overwhelming process of learning the technology for telemedicine, and (3) the desire for in-person or on-demand help with telemedicine.Substantial barriers exist for older adults in connecting with their health care team through telemedicine, particularly through video visits. The largest barriers include difficulty with technology or using the video visit platform, hearing difficulty, language barriers, and lack of desire to see providers virtually. Efforts to improve telemedicine access for older adults should take into account patient perspectives.

    View details for DOI 10.2196/34326

    View details for PubMedID 35438648

  • MRI Radiogenomics of Pediatric Medulloblastoma: A Multicenter Study. Radiology Zhang, M., Wong, S. W., Wright, J. N., Wagner, M. W., Toescu, S., Han, M., Tam, L. T., Zhou, Q., Ahmadian, S. S., Shpanskaya, K., Lummus, S., Lai, H., Eghbal, A., Radmanesh, A., Nemelka, J., Harward, S. 2., Malinzak, M., Laughlin, S., Perreault, S., Braun, K. R., Lober, R. M., Cho, Y. J., Ertl-Wagner, B., Ho, C. Y., Mankad, K., Vogel, H., Cheshier, S. H., Jacques, T. S., Aquilina, K., Fisher, P. G., Taylor, M., Poussaint, T., Vitanza, N. A., Grant, G. A., Pfister, S., Thompson, E., Jaju, A., Ramaswamy, V., Yeom, K. W. 2022: 212137

    Abstract

    Background Radiogenomics of pediatric medulloblastoma (MB) offers an opportunity for MB risk stratification, which may aid therapeutic decision making, family counseling, and selection of patient groups suitable for targeted genetic analysis. Purpose To develop machine learning strategies that identify the four clinically significant MB molecular subgroups. Materials and Methods In this retrospective study, consecutive pediatric patients with newly diagnosed MB at MRI at 12 international pediatric sites between July 1997 and May 2020 were identified. There were 1800 features extracted from T2- and contrast-enhanced T1-weighted preoperative MRI scans. A two-stage sequential classifier was designed-one that first identifies non-wingless (WNT) and non-sonic hedgehog (SHH) MB and then differentiates therapeutically relevant WNT from SHH. Further, a classifier that distinguishes high-risk group 3 from group 4 MB was developed. An independent, binary subgroup analysis was conducted to uncover radiomics features unique to infantile versus childhood SHH subgroups. The best-performing models from six candidate classifiers were selected, and performance was measured on holdout test sets. CIs were obtained by bootstrapping the test sets for 2000 random samples. Model accuracy score was compared with the no-information rate using the Wald test. Results The study cohort comprised 263 patients (mean age ± SD at diagnosis, 87 months ± 60; 166 boys). A two-stage classifier outperformed a single-stage multiclass classifier. The combined, sequential classifier achieved a microaveraged F1 score of 88% and a binary F1 score of 95% specifically for WNT. A group 3 versus group 4 classifier achieved an area under the receiver operating characteristic curve of 98%. Of the Image Biomarker Standardization Initiative features, texture and first-order intensity features were most contributory across the molecular subgroups. Conclusion An MRI-based machine learning decision path allowed identification of the four clinically relevant molecular pediatric medulloblastoma subgroups. © RSNA, 2022 Online supplemental material is available for this article. See also the editorial by Verschakelen in this issue.

    View details for DOI 10.1148/radiol.212137

    View details for PubMedID 35438562

  • Maladaptive myelination promotes generalized epilepsy progression Nature Neuroscience (accepted) Knowles, J. K. 2022
  • White Matter Properties of the Optic Pathway in Children with Neurofibromatosis Type 1 with and without Optic Pathway Gliomas Bruckert, L., Lerma-Usabiaga, G., Beres, S., McKenna, E., Tam, L., Yeom, K., Campen, C. WILEY. 2021: S25
  • Machine-Learning Approach to Differentiation of Benign and Malignant Peripheral Nerve Sheath Tumors: A Multicenter Study. Neurosurgery Zhang, M., Tong, E., Hamrick, F., Lee, E. H., Tam, L. T., Pendleton, C., Smith, B. W., Hug, N. F., Biswal, S., Seekins, J., Mattonen, S. A., Napel, S., Campen, C. J., Spinner, R. J., Yeom, K. W., Wilson, T. J., Mahan, M. A. 2021

    Abstract

    BACKGROUND: Clinicoradiologic differentiation between benign and malignant peripheral nerve sheath tumors (PNSTs) has important management implications.OBJECTIVE: To develop and evaluate machine-learning approaches to differentiate benign from malignant PNSTs.METHODS: We identified PNSTs treated at 3 institutions and extracted high-dimensional radiomics features from gadolinium-enhanced, T1-weighted magnetic resonance imaging (MRI) sequences. Training and test sets were selected randomly in a 70:30 ratio. A total of 900 image features were automatically extracted using the PyRadiomics package from Quantitative Imaging Feature Pipeline. Clinical data including age, sex, neurogenetic syndrome presence, spontaneous pain, and motor deficit were also incorporated. Features were selected using sparse regression analysis and retained features were further refined by gradient boost modeling to optimize the area under the curve (AUC) for diagnosis. We evaluated the performance of radiomics-based classifiers with and without clinical features and compared performance against human readers.RESULTS: A total of 95 malignant and 171 benign PNSTs were included. The final classifier model included 21 imaging and clinical features. Sensitivity, specificity, and AUC of 0.676, 0.882, and 0.845, respectively, were achieved on the test set. Using imaging and clinical features, human experts collectively achieved sensitivity, specificity, and AUC of 0.786, 0.431, and 0.624, respectively. The AUC of the classifier was statistically better than expert humans (P=.002). Expert humans were not statistically better than the no-information rate, whereas the classifier was (P=.001).CONCLUSION: Radiomics-based machine learning using routine MRI sequences and clinical features can aid in evaluation of PNSTs. Further improvement may be achieved by incorporating additional imaging sequences and clinical variables into future models.

    View details for DOI 10.1093/neuros/nyab212

    View details for PubMedID 34131749

  • Radiomic Signatures of Posterior Fossa Ependymoma: Molecular Subgroups and Risk Profiles. Neuro-oncology Zhang, M., Wang, E., Yecies, D., Tam, L. T., Han, M., Toescu, S., Wright, J. N., Altinmakas, E., Chen, E., Radmanesh, A., Nemelka, J., Oztekin, O., Wagner, M. W., Lober, R. M., Ertl-Wagner, B., Ho, C. Y., Mankad, K., Vitanza, N. A., Cheshier, S. H., Jacques, T. S., Fisher, P. G., Aquilina, K., Said, M., Jaju, A., Pfister, S., Taylor, M. D., Grant, G. A., Mattonen, S., Ramaswamy, V., Yeom, K. W. 2021

    Abstract

    The risk profile for posterior fossa ependymoma (EP) depends on surgical and molecular status [Group A (PFA) versus Group B (PFB)]. While subtotal tumor resection is known to confer worse prognosis, MRI-based EP risk-profiling is unexplored. We aimed to apply machine learning strategies to link MRI-based biomarkers of high-risk EP and also to distinguish PFA from PFB.We extracted 1800 quantitative features from presurgical T2-weighted (T2-MRI) and gadolinium-enhanced T1-weighted (T1-MRI) imaging of 157 EP patients. We implemented nested cross-validation to identify features for risk score calculations and apply a Cox model for survival analysis. We conducted additional feature selection for PFA versus PFB and examined performance across three candidate classifiers.For all EP patients with GTR, we identified four T2-MRI-based features and stratified patients into high- and low-risk groups, with 5-year overall survival rates of 62% and 100%, respectively (p < 0.0001). Among presumed PFA patients with GTR, four T1-MRI and five T2-MRI features predicted divergence of high- and low-risk groups, with 5-year overall survival rates of 62.7% and 96.7%, respectively (p = 0.002). T1-MRI-based features showed the best performance distinguishing PFA from PFB with an AUC of 0.86.We present machine learning strategies to identify MRI phenotypes that distinguish PFA from PFB, as well as high- and low-risk PFA. We also describe quantitative image predictors of aggressive EP tumors that might assist risk-profiling after surgery. Future studies could examine translating radiomics as an adjunct to EP risk assessment when considering therapy strategies or trial candidacy.

    View details for DOI 10.1093/neuonc/noab272

    View details for PubMedID 34850171

  • Association of pediatric acute-onset neuropsychiatric syndrome with microstructural differences in brain regions detected via diffusion-weighted magnetic resonance imaging. JAMA Network Open Zheng, J., Frankovich, J., McKenna, E. S., Rowe, N. C., MacEachern, S. J., Ng, N. N., Tam, L. T., Moon, P. K., Gao, J., Thienemann, M., Forkert, N. D., Yeom, K. W. 2020; 3 (5): e204063
  • Cerebral volume and diffusion MRI changes in children with sensorineural hearing loss. NeuroImage: Clinical Moon, P. K., Qian, J. Z., McKenna, E., Xi, K., Rowe, N. C., Ng, N. N., Zheng, J., Tam, L. T., MacEachern, S. J., Ahmad, I., Cheng, A. G., Forkert, N. D., Yeom, K. W. 2020; 27: 102328
  • Electrical and synaptic integration of glioma into neural circuits. Nature Venkatesh, H. S., Morishita, W., Geraghty, A. C., Silverbush, D., Gillespie, S. M., Arzt, M., Tam, L. T., Espenel, C., Ponnuswami, A., Ni, L., Woo, P. J., Taylor, K. R., Agarwal, A., Regev, A., Brang, D., Vogel, H., Hervey-Jumper, S., Bergles, D. E., Suva, M. L., Malenka, R. C., Monje, M. 2019

    Abstract

    High-grade gliomas are lethal brain cancers whose progression is robustly regulated by neuronal activity. Activity-regulated release of growth factors promotes glioma growth, but this alone is insufficient to explain the effect that neuronal activity exerts on glioma progression. Here we show that neuron and glioma interactions include electrochemical communication through bona fide AMPA receptor-dependent neuron-glioma synapses. Neuronal activity also evokes non-synaptic activity-dependent potassium currents that are amplified by gap junction-mediated tumour interconnections, forming an electrically coupled network. Depolarization of glioma membranes assessed by in vivo optogenetics promotes proliferation, whereas pharmacologically or genetically blocking electrochemical signalling inhibits the growth of glioma xenografts and extends mouse survival. Emphasizing the positive feedback mechanisms by which gliomas increase neuronal excitability and thus activity-regulated glioma growth, human intraoperative electrocorticography demonstrates increased cortical excitability in the glioma-infiltrated brain. Together, these findings indicate that synaptic and electrical integration into neural circuits promotes glioma progression.

    View details for DOI 10.1038/s41586-019-1563-y

    View details for PubMedID 31534222

  • Methotrexate Chemotherapy Induces Persistent Tri-glial Dysregulation that Underlies Chemotherapy-Related Cognitive Impairment CELL Gibson, E. M., Nagaraja, S., Ocampo, A., Tam, L. T., Wood, L. S., Pallegar, P. N., Greene, J. J., Geraghty, A. C., Goldstein, A. K., Ni, L., Woo, P. J., Barres, B. A., Liddelow, S., Vogel, H., Monje, M. 2019; 176 (1-2): 43-+
  • QUANTIFICATION OF SHUNT VOLUME THROUGH VENTRICULAR SEPTAL DEFECT BY REAL-TIME 3-D COLOR DOPPLER ECHOCARDIOGRAPHY: AN IN VITRO STUDY ULTRASOUND IN MEDICINE AND BIOLOGY Zhu, M., Ashraf, M., Tam, L., Streiff, C., Kimura, S., Shimada, E., Sahn, D. J. 2016; 42 (5): 1193-1200

    Abstract

    Quantification of shunt volume is important for ventricular septal defects (VSDs). The aim of the in vitro study described here was to test the feasibility of using real-time 3-D color Doppler echocardiography (RT3-D-CDE) to quantify shunt volume through a modeled VSD. Eight porcine heart phantoms with VSDs ranging in diameter from 3 to 25 mm were studied. Each phantom was passively driven at five different stroke volumes from 30 to 70 mL and two stroke rates, 60 and 120 strokes/min. RT3-D-CDE full volumes were obtained at color Doppler volume rates of 15, 20 and 27 volumes/s. Shunt flow derived from RT3-D-CDE was linearly correlated with pump-driven stroke volume (R = 0.982). RT3-D-CDE-derived shunt volumes from three color Doppler flow rate settings and two stroke rate acquisitions did not differ (p > 0.05). The use of RT3-D-CDE to determine shunt volume though VSDs is feasible. Different color volume rates/heart rates under clinically/physiologically relevant range have no effect on VSD 3-D shunt volume determination.

    View details for DOI 10.1016/j.ultrasmedbio.2015.12.017

    View details for Web of Science ID 000373385300017

    View details for PubMedID 26850842