Page 76 - Read Online

Page 8 of 11 Elefante et al. Mini-invasive Surg 2021;5:7 I http://dx.doi.org/10.20517/2574-1225.2020.102

molecular imaging in meningiomas, even if the utility of SSTR II imaging needs more validation to be
[65]
confirmed .

FUTURE DIRECTIONS
Radiomics is an emerging field of research that extracts many features from medical images. There are two
categories of features, which can be extracted from the region of interest after the lesion segmentation,
semantic and agnostic ones. In detail, semantic features are commonly used in the radiology lexicon to
describe a lesion (e.g., shape, location, etc.), but in the radiomics field they are quantified through computer
assistance. On the other hand, diagnostic features describe lesion heterogeneity using quantitative
descriptors. They include first-, second-, or higher-order statistics. First-order statistical outputs consist
of the grey level histogram analysis of the lesion’s voxels. Second-order statistics are those obtained from
texture analysis. They describe relationships between voxels considering their contrast values. Finally,
higher-order statistics are obtained imposing filters to extract definite image patterns, such as fractal
[66]
analyses, wavelets, or Laplacian transforms of Gaussian bandpass filters . Radiomics can be coupled
with artificial intelligence, which employs algorithms to allow computers to learn directly from the data
and make predictions on unseen datasets, because of its better capability of managing this volume of data
compared to traditional statistics . In the study of meningiomas, radiomics and artificial intelligence
[67]
have shown promise in preoperative evaluation, recurrence and outcome prediction, and radiation
treatment planning. Preoperative prediction of the meningioma grade is important because it influences
[68]
the treatment strategy. Park et al. obtained an accuracy of 89.7% for the prediction of meningioma grades
using MR conventional and diffusion tensor imaging with a radiomics and machine learning approach;
furthermore, various texture parameters differed significantly between fibroblastic and non-fibroblastic
benign meningiomas. Volumetric assessment of meningiomas is also highly relevant for therapy planning
[69]
and monitoring. Using a multiparametric deep-learning model on routine MRI data, Laukamp et al.
investigated its performance in automated detection and segmentation of meningiomas in comparison to
manual segmentations, obtaining a strong correlation despite diverse scanner data. Moreover, prognostic
models based on clinical, radiologic, and radiomic feature have been investigated to preoperatively identify
meningiomas at risk for poor outcomes. In this setting, preoperative radiologic and radiomic features such
as apparent diffusion coefficient and sphericity have proved effective in predicting local failure and overall
[70]
survival in these patients . MR radiomics has also been implemented to predict early progression or
recurrence, which characterize a subset of skull base meningiomas, achieving good results (accuracy 90%) .
[71]
Finally, radiomics has proved useful in the definition of radiotherapy target volume, which represents a
critical step in treatment planning, in order to improve the texture-based differentiation of tumor from
[72]
edema and to differentiate vasogenic from tumor cell infiltration edema .
CONCLUSION
Although generally easily identified on the basis of some pictorial neuroimaging features, meningiomas
can raise some concerns in terms of tissue characterization and treatment selection. In particular, surgery
largely relies on MRI and CT scans examination, as the type of therapeutic approach can vary depending
on tumor size and location. Modern imaging tools are helpful in identifying more aggressive histological
behavior, defining vessel and brain involvement, and evaluating the need for adjuvant therapies; at the same
time, emerging post-processing techniques can enhance tumor biology tracking and response to therapy
prediction. All these imaging-derived data coupled together may allow for optimal therapeutic planning
and tailored longitudinal follow-up, based on both patient and tumor fingerprinting.

DECLARATIONS
Authors’ contributions
Made substantial contributions according to ICMJE criteria: Elefante A, Russo C, Di Stasi M, Vola E, Ugga L,
Tortora F, De Divitiis O
Conception and design: Elefante A