NeuroSurg Predict Outcome risk calculators for neurosurgery. Research use only. All calculators
Research decision support

Surgical risk prediction for skull-base meningioma

Conti-Bendor-SBM estimates a patient's post-operative risk across five clinically meaningful outcomes from routinely available pre-operative variables. Complete one structured form and receive model-derived probabilities to support, not replace, clinical judgement.

Conti–Bendor Surgical Risk Predictor for Skull-Base Meningioma Surgery

5
Predicted outcomes
743
Patients in cohort
4
Tertiary centres
41
Pre-operative inputs
How it works

Transparent, temporally validated, clinically grounded

Predictions draw on a multicentre registry of operated skull-base meningiomas. For every outcome, several model families were trained and compared, and the single best-performing model was frozen for use in this tool.

1

Structured input

You enter 41 pre-operative variables through one guided form: demographics, comorbidity, functional status, clinical presentation and radiological characteristics.

2

Frozen best model

Each outcome uses the model that performed best under temporal validation. The operating threshold is the one frozen during that validation.

3

Probability & band

The model returns an individual probability, placed into an outcome-specific Low / Moderate / High band shown alongside the cohort base rate.

4

Clinical interpretation

Results flag cases for focused review and counselling. They never replace the complete clinical picture of the individual patient.

For research and investigational use only. These estimates are derived from statistical patterns in historical data and may support patient counselling, shared decision-making, and perioperative planning when interpreted within the individual patient's complete clinical context. As the models have not yet been prospectively validated or locally recalibrated, estimates should be interpreted with caution and should complement, not replace, clinical judgement.
What the tool predicts

Five post-operative outcomes

Each estimate is produced by a model trained and validated specifically for that outcome. Open "Strongest predictors" to see the factors most associated with each outcome across the cohort.

Interactive tool

Risk Calculator

Describe the patient below and calculate the five risk estimates. All fields are pre-filled with a default; change only what applies to this patient.

For research and investigational use only. Not a medical device. Estimates complement, and never replace, clinical judgement. Nothing you enter is stored: values are processed in memory and discarded. Our hosting provider sees your IP address briefly for routing and abuse protection; we never record it. Complete every field before calculating.

5 outcomes are computed together, each with its pinned model.

About

Cohort, validation & performance

Conti-Bendor-SBM was developed on a multicentre registry of operated skull-base meningiomas, using a temporal validation design so that models are evaluated on patients treated after those they were trained on.

743
Patients
4
Tertiary centres
3
Countries
2010–2025
Study period

Authors & contributions

The project was conceived and scientifically directed by the neurosurgical team, under the vision and senior leadership of Prof. Alfredo Conti, together with Dr. Noa Ben Dor. The data-science team provided the methodological and technical expertise to develop the machine-learning models and translate them into the web-based predictive calculator.

Clinical Vision & Scientific Leadership
Alfredo Conti
Neurosurgery
Associate Professor, Department of Biomedical and Neuromotor Sciences (DIBINEM), University of Bologna; IRCCS Istituto delle Scienze Neurologiche di Bologna
Noa Ben Dor
Neurosurgery Research & Clinical Innovation
Department of Neurosurgery, IRCCS ISNB. Conceived and developed the project, coordinated the research and modelling work, and led its translation into the predictive calculator.
Machine Learning & Technical Development
Davide Evangelista
Machine learning & implementation
Junior Assistant Professor, Department of Computer Science and Engineering (DISI), University of Bologna
Elena Loli Piccolomini
Methodology
Associate Professor of Numerical Analysis, Department of Computer Science and Engineering (DISI), University of Bologna
Gal Ziv
External collaborator, Computer Science
Prototyped the predictive pipeline and calculator interface

For further details on the authors, please see the publication or search online.

Validation design

Models were trained and internally compared under a temporal split: the earlier patients form the training set and the most recent patients form a held-out temporal test set. Operating thresholds were frozen on a separate validation split before the test set was seen, and are read at runtime by this calculator rather than tuned interactively.

Per-outcome performance

OutcomeModelAUROCThreshold

30-Day Mortality and Severe Complication (Ibañez ≥ 3) were evaluated on very few test events, so their probabilities are unreliable and are presented with a caveat; the associated factors are the more defensible content on those cards.

Risk bands

Three bands per outcome. The lower boundary is that outcome's frozen operating threshold; the upper is the 90th percentile of predicted probability in the temporal test set. By construction, High is the top ~10% of cases and Low is the below-threshold group; most cases land in Moderate, which is the correct behaviour for these models.

Ethics & governance

Ethics approval: CE-AVEC, Bologna (protocol 19029). For research and investigational use only; not a medical device and not for autonomous clinical decision-making.

Contact & citation

Contact: the Conti-Bendor-SBM research team (details to be added on publication).
Citation: full citation will be provided once the study is published.