Ref H10
Computational Service
Confidence Reported
Scoped After Feasibility Review
Service Overview
Radiomics and pathomics extract quantitative features from images that the eye does not reliably measure, then relate them to outcome, diagnosis or treatment response. We handle the full chain: image preprocessing and standardisation, segmentation, feature extraction to reproducible standards, feature selection and predictive modelling with validation that holds up. Because radiomics is notoriously sensitive to acquisition settings, harmonisation and stability filtering are treated as core work rather than optional refinements.
Key Features & Advantages
Reproducible Feature Definitions
Feature extraction follows standardised definitions with documented preprocessing, so results are comparable across studies rather than specific to one pipeline.
Acquisition Variation Handled
Scanner and protocol effects are assessed and harmonised, and unstable features are removed, because ignoring this step is the most common reason radiomic models fail externally.
Validation Designed Against Overfitting
Nested cross-validation and held-out or external testing are used throughout; performance is reported on genuinely unseen data with confidence intervals.
Interpretability Included
Models are delivered with feature importance and explanation analysis, so the clinical collaborators can judge whether the signal is plausible or an artefact.
Where This Service Is Used
1Building imaging-based prediction models for diagnosis, prognosis or treatment response
2Quantifying tumour heterogeneity from CT, MRI or PET imaging
3Analysing whole-slide pathology images for morphological and cellular features
4Processing functional imaging data including fMRI and fNIRS for group comparisons
5Automating segmentation and measurement across a large imaging cohort
Methods and Techniques Applied
1Image preprocessing: resampling, intensity normalisation, bias field correction and denoising as appropriate
2Manual, semi-automatic or deep-learning segmentation of regions of interest with reproducibility assessment
3Feature extraction covering intensity, shape, texture and wavelet-transformed descriptors
4Stability filtering across scanners, protocols and observers, with harmonisation applied where needed
5Feature selection and predictive modelling with nested cross-validation and independent testing
6Interpretability analysis, calibration assessment and, for clinical models, decision curve analysis
What We Need From You
1Imaging data in DICOM, NIfTI or whole-slide format with acquisition parameters
2Segmentations or annotations if available; otherwise segmentation is part of the scope
3Clinical outcome or label data with the endpoint definition and follow-up information
4Details of scanners, sites and protocols so that batch structure can be assessed properly
What You Receive
1Extracted feature matrices with the full extraction configuration recorded
2Segmentation masks and quality assessment
3Trained models with validated performance metrics and confidence intervals
4Feature importance, calibration and decision curve analyses
5Analysis code, model files and a methods section written for publication
What We Compute and Report
1Standardised feature matrices covering intensity, shape, texture and transformed descriptors
2Feature stability statistics across scanners, protocols and observers
3Model performance with discrimination, calibration and confidence intervals on unseen data
4Decision curve analysis indicating whether the model would improve real decisions
5Feature importance and per-case explanation output
Questions We Are Asked Before Scoping
How many cases do we need?
▸It depends on event rate and effect size, but radiomic models built on fewer than roughly a hundred cases rarely generalise. We give a realistic assessment during scoping rather than after the analysis.
Will the model work at another hospital?
▸Only if it was designed and validated for that, which requires multi-site data and harmonisation. Single-site models routinely lose substantial performance externally and we report that risk explicitly.
Are these models clinically usable?
▸No. Deliverables are research-use only. Clinical deployment requires regulatory clearance, prospective validation and a quality system, none of which are part of this service.
Interpretation and Limitations
Computational predictions prioritise hypotheses; they do not independently establish binding, biological activity, therapeutic efficacy, safety or clinical validity. Performance depends on input quality, target class, the structural evidence available and the applicability domain of the selected methods. Scores used for ranking are comparative devices and should not be read as measured quantities. All deliverables are for research use only and are not intended for diagnostic or therapeutic use.
Is This the Right Service for You?
Choose this when
Imaging features need relating to outcome, diagnosis or treatment response
Segmentation and quantification are required across a large cohort
A model must be built to a standard that reviewers and clinicians will accept
Consider another service when
Molecular rather than imaging features are available
Clinical Prediction Modeling Spatial molecular data is the modality in hand
Spatial Transcriptomics Analysis What Goes Wrong, and How We Handle It
These are the failure modes that produce confident-looking results which do not survive the next experiment. Each one is designed against rather than discovered afterwards.
Features that encode scanner and protocol rather than biology
Stability across acquisition settings is tested and unstable features are removed before modelling.
Performance reported on data the model has effectively seen
Nested cross-validation with held-out or external testing is used, and performance is quoted with confidence intervals.
A strong area-under-curve presented as clinical usefulness
Calibration and decision curve analysis are reported, since a discriminating model can still be useless at the relevant threshold.
A Worked Example
A radiomic signature that did not survive the second site
The situation
A single-centre model reported excellent discrimination and was being prepared for prospective use.
What the work involved
Feature stability was tested across scanners, harmonisation was applied, and the model was refitted and tested on data from an independent site.
What came back
Roughly 40% of the original features were scanner-dependent. After removal and harmonisation, performance fell substantially but transferred across sites — an honest baseline for the prospective plan.
Illustrative of the shape this work takes. Details are generalised, no client or target is identified, and scope and outcome vary with the system in question.
Service Workflow
1
Scientific Consultation
We discuss the objective, the decision it supports and the data you hold. Nothing is quoted before we understand what the result has to answer.
2
Feasibility Review
We check that the data, target class and available structural evidence can support the proposed approach, and say plainly where they cannot.
3
Technical Proposal
Scope, methods, deliverables, timeline and price are confirmed in writing, together with the limitations we expect to report.
4
Computation and QC
The analysis runs on pinned tool versions in a containerised environment, with quality checkpoints and method comparison where that is informative.
5
Report and Follow-Up
You receive the ranked results, the documented methods, the limitations and a proposed validation plan, plus support in interpreting them.
Service Process
1
Online Consultation
›
2
Solution Matching
›
3
Service Contract
›
4
Project Execution
›
5
Data Delivery
Specifications & Turnaround
Note: ICDMO delivers computational (in silico) services. Results prioritise hypotheses for testing; they do not by themselves establish binding, biological activity, efficacy or safety. All deliverables are for research use only and are not intended for diagnostic or therapeutic use. Downstream experimental work is scoped separately and executed through qualified partners.
Get a Custom Quote
Our scientific team responds within 24 hours with a detailed technical proposal and pricing tailored to your project.
Contact Us Online ConsultationStandard Deliverables
✓Sequence-verified final product
✓Full QC documentation
✓Certificate of Analysis (CoA)
✓Electronic data files
✓Project completion report
Why ICDMO?
Computational CDMO Focus
Reproducible, Versioned Pipelines
Confidence & Limitations Reported
Experiment-Aware Recommendations
Response Within One Business Day