Status
Conditions
Study type
Funder types
Identifiers
About
This observational study (STRATUM-OS) aims to collect the necessary data from a cohort of patients with planned surgery for suspected intra-axial malignant brain tumours (both primary and secondary) following the standard surgical procedure established in current clinical protocols. These data will serve two primary purposes:
i) To gather multimodal data (pre, intra and postoperative) essential for the development and technical validation of a 3D decision support tool for brain surgery guidance and diagnostics integrating augmented reality and multimodal data processing powered by artificial intelligence algorithms (called STRATUM tool);
ii) To collect outcome measures that will facilitate a subsequent comparative study (a non-randomized controlled clinical trial, called STRATUM-NRCCT) assessing the standard procedure alone versus the standard procedure augmented with the STRATUM tool. Patients from STRATUM-OS will act as a historical control group in the subsequent historically controlled clinical trial (STRATUM-NRCCT), which will be performed once STRATUM-OS has been completed.
In STRATUM-OS patients will receive standard care as per established clinical protocols, with no modification to their treatment. However, patients will be asked to grant access to their clinical information, complete questionnaires, and provide relevant pre, intra and postoperative information related to the surgical intervention. Data will be gathered from multiple sources, such as the Electronic Health Records (EHR), patient completed questionnaires, interviews, and reports from healthcare professionals involved in the surgical procedure. Additionally, intraoperative data will be collected from the different devices in the operating room.
Full description
CONTEXT OF EUROPEAN STRATUM PROJECT
Integrated digital diagnostics can support complex surgeries in many anatomies where brain tumour surgery is one of the most complex cases. Neurosurgeons face several challenges during brain tumour surgeries, such as critical tissue and brain tumour margins differentiation or the interpretation of large amount of data available provided by several independent devices.
To address these challenges, STRATUM aims to develop a 3D decision support tool for brain surgery guidance and diagnostics integrating augmented reality and multimodal data processing powered by artificial intelligence (AI) algorithms. This tool will function as a point-of-care computing system and will be developed using a co-creation methodology that actively involves end-users and other stakeholders. The STRATUM tool will include hyperspectral (HS) imaging (HSI) as an emerging imaging modality in the medical field to enhance intraoperative guidance and diagnosis during the neurosurgical procedures.9 Previous works from several members of the STRATUM consortium in different research projects (HELICoiD, ITHaCA, and NEMESIS-3D-CM) have demonstrated, as a proof-of-concept, that this technology is suitable for the intraoperative identification and delineation of brain tumours in real-time. Additionally, the tool is expected to provide a real-time deformation of the magnetic resonance imaging (MRI) within the exposed brain surface for brain-shift compensation during surgery. This will be performed by using advanced mathematical models in combination with the intraoperative multimodal data (HSI, depth information and standard surgical microscope imaging) captured by the STRATUM tool.
The system will be developed and clinically evaluated in three main stages. In Stage 1, a customized multimodal data acquisition system was developed to be used in the observational study of Stage 2 (STRATUM-OS), which will focused on the multimodal data imaging collection to support the development and technical validation of the STRATUM tool. In Stage 3 the tool will undergo clinical validation through a subsequent, non-randomized, historically controlled, clinical trial (STRATUM-NRCCT). The historic control in STRATUM-NRCCT will be the subjects recruited in STRATUM-OS.
Overall, the STRATUM project aims to:
i) optimize the integration and processing of existing and emerging data sources, facilitating timely, efficient and accurate surgical decision-making; ii) maximize tumour resection while minimizing the risk of neurological deficits; iii) reduce anaesthesia duration and related risks; iv) decrease waste associated with repeated pathology analysis; and v) optimize healthcare resource utilization.
STRATUM OBSERVATIONAL STUDY
STRATUM-OS is an international multicentre, prospective, open, observational cohort study, with a follow-up duration of 6 months, in which the data generated from brain tumour surgeries, including a wide range of intra-axial tumour types, will be collected to meet the objectives of the study. In STRATUM-OS patients will receive standard care as per established clinical protocols, with no modification to their treatment. However, patients will be asked to grant access to their clinical information, complete questionnaires, and provide relevant pre, intra and postoperative information related to the surgical intervention. STRATUM-OS is planned for a duration of 28 months divided in: i) a pre-recruitment period of 2 months for the installation of and surgeon training on the acquisition system, ii) a recruitment period of 20 months, and iii) follow-up period of 6 months, including one month for the integration and technical validation of the fully-working STRATUM tool. We anticipate that 320 consecutive patients can be recruited during this study in the 3 clinical sites. The protocol has been drafted in accordance with the Standardised Protocol Items: Recommendations for Observational Studies (SPIROS) statement
The general objective of STRATUM-OS is to collect the necessary data from a cohort of patients affected by intra-axial brain tumours with the standard surgical procedure established in current clinical protocols. STRATUM-OS will pursue the following main objectives:
To collect pre-stored and in-situ multimodal data for the development of an intraoperative 3D decision support tool for brain surgery guidance and diagnostics in real-time leveraging AI-based multimodal data processing (STRATUM tool).
To technically validate the STRATUM tool, aiming for (1) the intraoperative distinction between tumour and non-tumour areas in the exposed brain surface and (2) the identification of contrast-enhancing tumour (CET) or non-contrast-enhancing tumour (nCET/FLAIR-positive) regions in MRI, through AI-driven processing.
To compile a historical control group dataset including patient clinical data, health outcomes, surgical and tumour characteristics, and hospital resource utilization and costs. This dataset will be used in the subsequent non-randomized controlled clinical trial (STRATUM-NRCCT), to assess the safety, effectiveness and cost-effectiveness of the STRATUM tool in brain tumours surgery.
SETTING AND RECRUITMENT
Adult participants (≥ 18 years) with an intra-axial brain tumour will be eligible for inclusion. Recruitment will follow a consecutive enrolment process, selecting subjects who meet all the inclusion criteria and none of the exclusion criteria at the 3 participating clinical institutions: Hospital Universitario de Gran Canaria Doctor Negrín (Las Palmas de Gran Canaria, Spain), Karolinska University Hospital (Solna, Sweden) and Hospital Universitario 12 de Octubre (Madrid, Spain). Patients will be invited to participate and will be required to sign a written, informed consent form prior to inclusion in the study. They will continue to receive care at their originally assigned medical centre, with no patient transfers between institutions. Members of the research team at each hospital site will introduce the study to subjects who will receive written information describing the study. Researchers will discuss the study details with participants ensuring they have a thorough understanding before making a decision. Participants will have the opportunity to engage in an informed discussion with their physician before consenting. Written informed consent will be obtained from participants or, when applicable, from their designed tutor or legal representatives.
DATA COLLECTION
The data collection procedure will include data extracted from the EHR of the patient, self-reported questionnaires, information collected from the different professionals involved in the neurosurgical workflow, recorded through an electronic Case Report Form (eCRF), and data collected intraoperatively using the STRATUM acquisition system along with detailed information about the surgery (using the eCRF). The STRATUM eCRF is built on the REDCap (Research Electronic Data Capture) platform and securely stored in an anonymized and standardized format within a secure repository at the Institute for Applied Microelectronics (IUMA) of the University of Las Palmas de Gran Canaria (ULPGC). All patient data will be assigned by a unique coded ID [identification] number linked to the subject to ensure pseudo-anonymization. Only the local clinical team will be aware of each participant's identity. A locally and securely managed document will link each study ID with the corresponding participant. The data collection procedure will be divided into three main phases.
Preoperative phase:
Patients who meet the inclusion criteria and none of the exclusion criteria and after giving consent to participate in the study will be identified by the Data Collector (DC) at each clinical site. The DC will extract preliminary information from the EHR and confirm the eligibility with the principal investigator at the site before surgery. Preoperative data, including tabular patient information and various preoperative imaging modalities, will be collected, anonymized and transcribed by the DC from several sources (EHR, self-reported questionnaires and interviews/questionnaires/reports from healthcare professionals involved in the neurosurgical workflows). These data will be entered into the STRATUM eCRF.
Intraoperative phase:
During surgery, the operating surgeon will be assisted by the DC in carrying out the following tasks:
Postoperative phase:
Post-operative data will be collected by the DC from the EHR at one, three, and six months after surgery. The same anonymization protocol will be applied, ensuring that all multimodal data is securely stored on the STRATUM server, while tabular data is entered into the STRATUM eCRF.
Once data collection is completed for each patient, the Study Monitor, who is independent of the research team, will review the database for errors and missing data, ensuring data quality. If necessary, the Study Monitor will collaborate with the site DC to resolve errors or discrepancies between the eCRF and the primary source data (e.g., EHR and questionnaires) and make direct revisions in the eCRF. In case of a participant ceases participation in the study or is lost to follow-up, the anonymized data generated until that moment will be employed, if possible, for the technical validation, but the observation will be excluded from the analyses where the missing data is necessary to compute the outcomes in the subsequent clinical trial.
STUDY MEASURE CATEGORIES
The collected measures will span from patient enrolment to the end of a 6-month follow-up period after surgery and will be categorized in the following main domains:
SAMPLE SIZE
In total, it is estimated that 26 patients will undergo brain surgery per month across the three clinical sites. Of these, approximately 70% of patients (~18 patients/month) are expected to meet the inclusion criteria, satisfy none of the exclusion criteria, and provide the informed consent for study participation. A 90% success rate in obtaining usable samples is anticipated (~16 patients/month). Consequently, STRATUM-OS is expected to collect data from 320 consecutive patients over 20 months of recruitment, followed by a 6-month follow-up period (total duration: 28 months). Given an average of 4.5 samples per patient (accounting for potential sample loss during pathological analysis), a total of approximately 1,440 tissue samples are expected to be collected.
These estimations have been obtained based on previous experiences from the project partners. Particularly, from the HELICoiD and ITHaCA projects in which the same HS acquisition system was employed, a total of 85 HS images were obtained from 41 different subjects captured in three data acquisition campaigns at the Hospital Universitario de Gran Canaria Dr. Negrín, covering a 24-months period in total. From this dataset, 28% of HS images were excluded (17% of subjects), resulting in 61 HS images from 34 eligible subjects. Additionally, the study conducted at the Hospital Universitario 12 de Octubre within the NEMESIS-3D-CM project was able to obtain a multimodal dataset composed by HS images from 193 different subjects, also in a 24-months period.
Although patient-level data are important for clinical context and for using them as historical controls in the subsequent STRATUM-NRCCT study, in this study the primary unit of analysis for the main outcome measure is the individual tissue sample, which will be histologically classified as "tumour" or "non-tumour" and serve as the reference standard for technical validation. Therefore, the effective sample size for the statistical analysis is determined by the number of validated tissue samples. This volume of data is expected to provide sufficient statistical power to estimate key diagnostic performance metrics of the STRATUM Tool-such as sensitivity, specificity, and predictive values-with acceptable precision.
A preliminary evaluation of inclusion rates will be conducted after the first 3 months of recruitment, to identify potential barriers to enrolment. If necessary, corrective measures will be implemented to ensure that the study reaches its target sample size.
DATA PARTITION FOR TECHNICAL VALIDATION
In AI-based applications, data partitioning is the process of dividing a dataset into several subsets (e.g., training, validation and test sets). This process is crucial to evaluate and validate the performance of developed AI models, ensuring that models are validated and tested using unseen data for model training. This is highly important especially in medical applications, where data from different subjects must be in independent sets. This allows a more accurate assessment of the model performance, avoiding overfitting and obtaining more generalized models for unseen data/subjects.
In this study, we plan to utilize the initial 70% of recruited patients to train the AI algorithm (n=224 patients). The subsequent 10% will be used for cross-validation (n=32), and the final 20% will serve to test the model (n=64). While data partitioning is performed at the patient level to avoid information leakage, the actual analysis will be conducted at the tissue sample level, using labels validated by histopathology to assess diagnostic performance at the tissue sample level.
STATISTICAL METHODS
To ensure a robust and unbiased evaluation of the STRATUM Tool, the dataset will be partitioned at the patient level into three non-overlapping subsets: 70% for model training, 10% for internal validation, and 20% for final testing. This partitioning strategy is intended to prevent data leakage across subsets, ensuring that all tissue samples from the same patient are assigned to the same group. Observations from participants with missing data may be excluded from analyses.
The training set will be used to build the AI model by learning patterns from intraoperative imaging data paired with corresponding histopathological or radiological labels. The validation set will be used during model development to fine-tune hyperparameters and optimize training procedures (e.g., early stopping, learning rate adjustments, regularization). No metrics will be formally reported from the validation set.
The test set will be completely isolated during model development and used exclusively to compute final performance metrics. These will include accuracy, sensitivity, specificity, precision, F1-score, ROC curve analysis, the Jaccard Index, and the Dice-Sørensen coefficient, as appropriate to the outcome type (classification or segmentation). Diagnostic performance will be assessed for two main tasks: (1) the distinction between tumour and non-tumour tissue samples, using definitive histopathological analysis as the reference standard; and (2) the identification of CET and nCET (FLAIR-positive) regions in MRI, using histopathological and radiological reports as the reference.
Subgroup analyses will be performed for the most common histological tumour types, defined as those with at least 25 patients included in the test set. All subgroup evaluations will be conducted exclusively within the test set to ensure unbiased estimation. Due to the subsample nature of the investigation, performance estimates may have wide confidence intervals, particularly for less frequent tumour subtypes.
ETHICS AND DISSEMINATION
The study will adhere to the ethical principles for medical research involving human subjects established in the Declaration of Helsinki and the Good Clinical Practice Guidelines. According to our previous experience, the use of HSI has not demonstrated any safety or tolerability concerns in surgical procedures.
Multimodal data will be captured by expert neurosurgeons using the STRATUM acquisition system designed, produced, and installed at each clinical site (at the time of surgery, no real-time results on tissue classification will be displayed to physicians, except for the standard frozen section histopathological diagnostic information they usually receive). The STRATUM acquisition system will not alter the surgical procedure, apart for the data collection process (estimated to be ~10 min during the entire surgery with no expected negative effects for the patient). Captured data will not influence or modify the neurosurgical plan. As part of the standard procedure, tumour tissue samples and adjacent tissue samples (suspected to be tumour) will be collected for pathological analysis. These samples will serve as golden standard for the algorithm development. This sample collection will not interfere with the intervention, histopathological analysis, or the intraoperative decision-making. Pathologists will have no access to the STRATUM tool results prior to their independent analysis, even during validation and testing phases after initial training.
Patient confidentiality and data security will be managed in compliance with General Data Protection Regulations and relevant national and European legislations as per local and national ethical approvals. All study-related information will be securely stored at the clinical sites. All local databases will be protected by password restricted access systems. Forms, lists, logbooks, appointment books, and any other listings that link participant ID numbers to other identifying information will be stored in a separated, locked restricted-access areas. Only authorized personnel, including researchers involved in the STRATUM project, the sponsor or designated representatives, the Ethics Committee, and relevant health authorities will have access to this data.
All data and biological samples collected during STRATUM-OS will be used exclusively for the development and technical validation of the STRATUM tool, as well as for the creation of a historical control group for the subsequent non-randomized controlled clinical trial (STRATUM-NRCCT). This purpose is explicitly stated among the primary objectives of the present protocol. Therefore, no additional informed consent will be required for this use, as participants will be fully informed and provide consent to both components-technical validation and historical control generation-at the time of inclusion. Any future use of data or samples beyond the scope of this protocol will require prior approval from the relevant Ethics Committees and, where applicable, new participant consent. Participants will have the right to access, rectify, delete, limit the processing, portability and opposition of their data by contacting the principal investigator of the project in each clinical site.
The results of this study will be published in open access journals, regardless of whether the findings are positive or negative. The study results will be shared with the participating physicians, referring clinicians, patients, and the broader medical and scientific community. Data (properly anonymized) will be stored in the secure STRATUM repository at the project coordination institution and, upon project completion, will be archived in trusted repositories, having their respective digital object identifiers (DOIs).
Enrollment
Sex
Ages
Volunteers
Inclusion criteria
Exclusion criteria
320 participants in 1 patient group
Loading...
Central trial contact
Himar Fabelo, PhD
Data sourced from clinicaltrials.gov
Clinical trials
Research sites
Resources
Legal