Brain Cancer Radio-Pathomics

Brain Cancer

Relevance

Glioblastomas (GBMs) are the most common central nervous system originating tumor, accounting for around 50% of new brain-originating cancer diagnoses. After abnormal findings on magnetic resonance imaging (MRI), surgical resection is conducted to debulk the tumor and provide tissue samples for pathological diagnosis. Gross total resection is often attempted. Patients are then typically treated with localized radiation treatment with concordant temozolomide therapy and may receive additional treatments such as anti-angiogenic agents (i.e., bevacizumab) or tumor treating fields (TTFields) therapy. Despite the therapeutic benefit of these treatments, patient prognosis is dismal, with one- and five-year survival rates of 41% and 6.6%, respectively. Critical to maximizing the efficacy of GBM treatment is accurately localizing the full extent of the tumor. Non-invasive tumor monitoring via MRI includes pre- (T1) and -post-Gd contrast (T1C) T1-weighted imaging, which is typically used to define the primary tumor mass, T2-weighted FLAIR imaging, where hyperintense areas may reflect active tumor or edema, and apparent diffusion coefficient (ADC) images calculated from diffusion weighted imaging, where areas of restricted diffusion are thought to correlate with hypercellularity. However, autopsy studies have found areas of infiltrative tumor as far as 10 cm beyond the T1C and FLAIR margins and have shown little-to-no relationship between ADC values and cellularity when using tissue collected from beyond the suspected tumor region. Therefore, there is a pressing need to develop new imaging tools to more precisely localize treatment and monitor tumor progression to improve patient prognoses.

Our Work

Our neuro-oncology research focuses on aligning autopsy tissue samples from our brain bank to clinical MR imaging to develop artificial intelligence-based signatures of tumor presence. As conventional MRI often fails to show the full extent of tumor invasion, the tools we develop seek to use data science techniques to find microscopic pathological signatures that identify invisible tumor invasion and progression. By applying these mapping techniques to patient data, we are able to help clinicians direct treatment to the full extent of tumor presence and more precisely monitor disease progression.

 In addition to developing these models from our brain bank data, we have also used our maps on a range of additional datasets to better understand how unseen areas of tumor relate to different disease-related factors, such as how they respond to treatments such as bevacizumab and how they impact patient prognosis. Additionally, we have compared our maps to other forms of glioma imaging such as perfusion, diffusion, MR-spectroscopy, and positron emission tomography (PET) data to better understand the pathophysiology of non-enhancing tumor. We have had fruitful internal collaborations with clinicians to facilitate the translation of this technology into a robust clinical tool, as well as external and international collaborations to demonstrate our model’s capabilities on a wide range of collected data.

  • Bobholz SA, Lowman AK, Connelly JM, Duenweg SR, Winiarz A, Nath B, et al. Noninvasive Autopsy-Validated Tumor Probability Maps Identify Glioma Invasion Beyond Contrast Enhancement. Neurosurgery. 2024. Epub 20240319. doi: 10.1227/neu.0000000000002898. PubMed PMID: 38501824

  • Bobholz SA, Hoefs A, Hamburger J, Lowman AK, Winiarz A, Duenweg SR, et al. Radio-pathomic maps of glioblastoma identify phenotypes of non-enhancing tumor infiltration associated with bevacizumab treatment response. J Neurooncol. 2024 Apr;167(2):233-241. doi: 10.1007/s11060-024-04593-7. Epub 2024 Feb 19. Erratum in: J Neurooncol. 2024 Apr;167(2):243. doi: 10.1007/s11060-024-04641-2. PubMed PMID: 38372901; PubMed Central PMCID: PMC11024025

  • Bobholz SA, Lowman AK, Brehler M, Kyereme F, Duenweg SR, Sherman J, et al. Radio-Pathomic Maps of Cell Density Identify Brain Tumor Invasion beyond Traditional MRI-Defined Margins. AJNR Am J Neuroradiol. 2022;43(5):682-8. Epub 20220414. doi: 10.3174/ajnr.A7477. PubMed PMID: 35422419; PubMed Central PMCID: PMC9089258

  • Bobholz SA, Lowman AK, Barrington A, Brehler M, McGarry S, Cochran EJ, et al. Radiomic Features of Multiparametric MRI Present Stable Associations with Analogous Histological Features in Patients with Brain Cancer. Tomography. 2020 Jun;6(2):160-169. doi: 10.18383/j.tom.2019.00029. PMID: 32548292; PMCID: PMC7289245

  • McGarry SD, Hurrell SL, Kaczmarowski AL, Cochran EJ, Connelly J, Rand SD, et al. Magnetic Resonance Imaging-Based Radiomic Profiles Predict Patient Prognosis in Newly Diagnosed Glioblastoma Before Therapy. Tomography. 2016 Sep;2(3):223-228. doi: 10.18383/j.tom.2016.00250. PubMed PMID: 27774518; PubMed Central PMCID: PMC5074084.

Relevant Publications

Previous
Previous

Prostate Cancer

Next
Next

3D Printing