Non-invasive tumor probability maps developed using autopsy tissue identify novel areas of tumor beyond the imaging-defined margin

Abstract

Background: This study identified a clinically significant subset of glioma patients with tumor outside of contrast-enhancement present at autopsy, and subsequently developed a method for detecting non-enhancing tumor using radio-pathomic mapping. We tested the hypothesis that autopsy-based radio-pathomic tumor probability maps would be able to non-invasively identify areas of infiltrative tumor beyond traditional imaging signatures. Methods: A total of 159 tissue samples from 65 subjects were aligned to MRI acquired nearest to death for this study. Demographic and survival characteristics for patients with and without tumor beyond the contrast-enhancing margin were computed. An ensemble algorithm was used to predict pixelwise tumor presence from pathological annotations using segmented cellularity (Cell), extracellular fluid (ECF), and cytoplasm (Cyt) density as input (6 train/3 test subjects). A second level of ensemble algorithms were used to predict voxel-wise Cell, ECF, and Cyt on the full dataset (43 train/22 test subjects) using 5-by-5 voxel tiles from T1, T1+C, FLAIR, and ADC as input. The models were then combined to generate non-invasive whole brain maps of tumor probability. Results: Tumor outside of contrast was identified in 41.5% of patients, who showed worse survival outcomes (HR=3.90, p<0.001). Tumor probability maps reliably tracked non-enhancing tumor in the test set, external data collected pre-surgery, and longitudinal data to identify treatment-related changes and anticipate recurrence. Conclusion: This study developed a multi-stage model for mapping gliomas using autopsy tissue samples as ground truth, which was able to identify regions of tumor beyond traditional imaging signatures.

Radio-pathomic maps of glioblastoma identify phenotypes of non-enhancing tumor infiltration associated with bevacizumab treatment response

“…Processed MR images for each patient at each timepoint were then used to generate radio-pathomic maps of cell density and extracellular uid using a previously developed algorithm [23,32]. Brie y, large format autopsy samples were collected from areas of suspected tumor and non-tumor and aligned to clinical MRI near death [23,[36][37][38].…”

“…These maps were trained on 43 patients and tested on 22 held-out patients, showing high accuracy on internal test data and impressive generalizability to external data. The maps were further converted to tumor probability maps via an additional algorithm in the prior manuscript [32], but for this current study only cell density and ECF maps are used. Radiopathomic maps were generated for each imaging acquisition included in this study using this pre-trained model using a local Matlab toolbox, producing both maps in approximately 10 minutes per session.…”

“…Machine learning and deep learning studies using biopsy tissue as ground truth have exploited the increased computational e ciency of modern hardware to extract deep textural features from conventional imaging to improve the detection of occult tumor invasion, though these studies are limited to the scope and magnitude of surgically resectable tissue [30,31]. In recent studies, large format autopsy tissue from glioma patients have been aligned to MR images to develop a radiopathomic mapping tool that allows for non-invasive detection of cell density, extracellular uid density, cytoplasm density, and tumor probability [23,32]. By using a pathologically rich, post-treatment ground truth sampled beyond the presence of traditional imaging signatures, this model can detect previously invisible areas of invasion and distinguishes between areas of tumor and treatment effect.…”