The primary objective is to improve the accuracy to predict glioma genotype of our previously developed algorithm *PrognosAIs* by adding advanced MRI acquired before surgery. The secondary objective is to learn the relationship between MRI signal…
ID
Source
Brief title
Condition
- Nervous system neoplasms malignant and unspecified NEC
- Head and neck therapeutic procedures
Synonym
Research involving
Sponsors and support
Intervention
Outcome measures
Primary outcome
The primary study parameter is the diagnostic accuracy with which we can
predict glioma genotype, based on pre-operative advanced MRI techniques in
combination with our previously developed *PrognosAIs* algorithm.
Secondary outcome
The secondary parameter is the association between MRI signal and tissue
features (both at the cellular and molecular level).
Background summary
The iGENE2.0 project challenges the current diagnostic paradigm that entirely
relies on tissue for diagnosis and management of patients with a brain tumour,
with the ultimate aim to determine tumour genotype non-invasively from MRI
only.
Roughly 1,000 new cases of adult primary brain tumours are diagnosed each year
in the Netherlands. These are most often diffuse glioma, and outcome is largely
dismal. Prognosis and treatment decisions rely on tissue diagnosis of tumour
(geno)type and grade. Glioma typing is based on the presence or absence of
mutation in the isocitrate dehydrogenase (IDH) gene: IDH mutated (IDHmut)
respectively IDH wild type (IDHwt). In the presence of other specific, genetic
alterations, IDHwt tumours are classified as *glioblastoma, IDHwt* (GBM).
IDHmut tumours are further divided into those with versus without 1p/19q
co-deletion: *oligodendroglioma, IDHmut and 1p/19q co-deleted* respectively
*astrocytoma, IDHmut*.
Gross total surgical resection is treatment of first choice, but is often not
feasible, for instance due to tumour location or extent of infiltration in the
brain or due to patient frailty. In such cases, diagnostic biopsy is performed,
solely to obtain tissue for diagnosis. For patients, these are procedures with
high impact, have complication risks, and require hospitalisation. They are
therefore rarely performed in the setting of tumour recurrence, let alone at
intermediate time points. This means that in most patients there is no tumour
tissue available for assessing - altered - tumour characteristics to rationally
determine treatment switch or inclusion in trials along the course of their
disease. The non-invasive characterisation of brain tumours is thus clinically
highly relevant not only at first diagnosis - especially in patients not
qualifying for surgical resection - but particularly upon recurrence.
In our previous work, it has been shown that non-invasive genotypical
classification of adult-type glioma with only conventional MRI techniques can
be achieved using artificial intelligence (AI) with accuracies of ~85%. This is
not yet sufficient to replace surgical biopsy. One way to improve the
diagnostic accuracy is by computationally exploiting more advanced MRI data.
For example, in our previous work we found that with a combination of more
advanced MRI techniques such as diffusion weighted (DWI) and perfusion MRI
(PWI), reflecting cellularity and vascularisation respectively, IDHwt glioma
can be distinguished from IDHmut 1p/19q intact glioma. More recent advanced MRI
techniques such as Chemical Exchange Saturation Transfer (CEST) MRI, which
reflects cellular proliferation, have been shown to correlate well with a
tumour*s aggressiveness. This work underlies the hypothesis that MRI can be
sensitised to specific glioma tissue features. Also, current classification
predictions only scratch the surface of tumour biology, of which the
micro-environment is of particular interest in view of new treatment targets.
Computational analysis techniques based on radiological imaging (*radiomics*)
have taken a huge flight in the past decade and are increasingly successful in
disease detection, characterisation, and surveillance. Given the clinical
implications, a large body of work has focused on the prediction/correlation of
glioma genotype from/with imaging phenotypes: *radiogenomics*. Some exploratory
work indicates that MRI can also predict the tumour*s immune micro-environment
phenotype.
By leveraging the biological association between morphological tissue
properties and (advanced) MRI signal on the one hand, and genetic alterations
on the other, we expect to make clinically meaningful diagnostic predictions
based on pre-operative MRI alone, and thus eventually not requiring tumour
tissue currently only obtainable through invasive surgical procedures. In this
study, we assess whether we can improve such diagnostic predictions by
combining our previously developed algorithm *PrognosAIs* with advanced MRI.
Additionally, we aim to gain a better understanding of the relationship between
the (advanced) MRI signal on the one hand and the tumour tissue characteristics
(both on the cellular and molecular level) on the other hand. Such insights are
essential for the future development of AI models by indicating which of the
advanced MRI technique(s) have the highest potential and strongest biological
foundation to eventually make a fully MRI-based glioma diagnosis possible.
Study objective
The primary objective is to improve the accuracy to predict glioma genotype of
our previously developed algorithm *PrognosAIs* by adding advanced MRI acquired
before surgery. The secondary objective is to learn the relationship between
MRI signal formation and tissue characteristics, both at the cellular and
molecular level. Both these objectives contribute to guiding the future
development of artificial intelligence (AI) models, indicating which of the
advanced MRI technique(s) have the highest potential and strongest biological
foundation to eventually make a fully MRI-based glioma diagnosis possible.
Study design
In this prospective observational study, prospective data will be collected,
which will start once ethical approval is in place.
For the primary objective of this study, patients will be scanned with advanced
MRI techniques, from which quantitative parameters will be obtained which will
be investigated as predictors of brain tumour genotype. These predictors will
be combined with the prediction from our previously developed AI algorithm
*PrognosAIs*, through regression analyses. In parallel to this project,
*PrognosAIs* is further developed with retrospective data (MEC-2024-0211).
Future iterations of PrognosAIs will likely become available during the course
of this project and will similarly be combined with the advanced MRI
predictors.
Once included (after informed consent is given by the patients), patients will
be scanned at the department of Radiology & Nuclear Medicine at Erasmus MC with
a pre-operative MRI scan as part of the standard clinical care (30 minutes).
Additionally, advanced MRI sequences will be added to this scan session (max 30
minutes).
For the secondary objective of the project, targeted stereotactic biopsies will
be performed during surgery at specific tumour locations identified based on
the pre-operative MRI scans. This enables to establish a spatial relationship
between advanced MRI parameters and tissue characteristics, which is important
because of tumour heterogeneity both in terms of MRI signal and tumour
histology. However, since this procedure carries an - albeit limited -
additional risk to the patient while it is at the same time not essential for
the primary objective of the study, patients are given the opportunity to opt
out of this aspect of the study. In these cases, the location of the diagnostic
biopsies performed as part of the standard clinical routine will be recorded
during the surgery, such that a relationship between the clinically acquired
and assessed tissue sample and the MRI signal can still be investigated.
The collected tissue samples will - in addition to routine clinical analysis -
be subjected to (advanced) immunohistochemical staining and molecular analyses
such as next-generation sequencing (NGS), for tumour microenvironment mapping
respectively genetic profiling (e.g., IDH mutation status). These quantitative
tissue data will be correlated with the advanced MRI parameters, with the known
locations from the pre-operative scan.
Patients will be followed up for disease progression and survival in routine
clinical practice, and their quality of life and experience during the study
will be recorded. No additional observations beyond those in routine clinical
care will be conducted in the context of this study.
Study burden and risks
No additional benefits are associated with patient participation in this study.
However, patient participation could help the development of techniques capable
of diagnosing primary brain tumours without the need for invasive procedures,
therefore helping future patients.
One possible burden for participating patients is that they will be subjected
to longer MRI scanning times than usual, as aside from the conventional MRI
scans that patients with glioma undergo, they will be scanned with several
advanced MRI techniques. The total additional scan time is 30 minutes.
The surgeries that these patients will undergo (either for diagnostic biopsy or
a resection) are part of the standard medical procedure. However, in patients
opting to undergo targeted biopsies, the surgical time will be extended by a
maximum of 15-30 minutes depending on the intended procedure (diagnostic biopsy
or resection respectively). The additional targeted biopsies carry a - limited
- risk of haemorrhage (1.5%). These risks will be mitigated by careful
pre-operative planning with the aim to take all diagnostic and targeted
biopsies from within the same trajectory whenever possible.
Dr. Molewaterplein 40
Rotterdam 3015 GD
NL
Dr. Molewaterplein 40
Rotterdam 3015 GD
NL
Listed location countries
Age
Inclusion criteria
Participant must be an adult (18 years or older)
Participant must be scheduled for surgery (resection or biopsy) of a (presumed)
primary brain tumour
Participant must give written informed consent conform ICH-GCP
Exclusion criteria
Contra-indication for MRI (implanted metal parts, e.g. stents, vascular clips,
pacemakers, claustrophobia)
Inability to give consent
Having received/receiving chemotherapy for a brain tumour at time of MRI, prior
to surgery
Design
Recruitment
Followed up by the following (possibly more current) registration
No registrations found.
Other (possibly less up-to-date) registrations in this register
No registrations found.
In other registers
Register | ID |
---|---|
CCMO | NL87585.078.24 |