NeuroScanner Rare Diseases

NeuroScanner Rare Diseases: AI Clinical Decision Support for Early Screening of Pediatric Orphan Disorders

Executive Overview

NeuroScanner Rare Diseases (CDSS RD) is an enterprise clinical decision support system that screens de-identified electronic health records (EHR) to find children at high risk of rare genetic disorders. Rare diseases often go undiagnosed for years: the average path from the first clinical visit to a diagnosis is about seven years. NeuroScanner combines medical natural language processing, entity extraction and deep neural representation models to read histories, laboratory panels and family pedigrees in about 80 milliseconds per record, and ranks patients by risk so that geneticists can focus confirmatory tests on the top 0.5% of the highest-probability cohort. In validation it helps bring time to diagnosis down from about 7 years to about 2. Built on the AI-Systems Smart Clinic platform.

Neural & Technological Stack

• Medical NLP core: custom clinical named entity recognition, rule-based lexicons, spelling correction and lemmatisation for physician epicrises, phenotypic signs, pedigree data and laboratory biomarkers

• Representation learning: PyTorch autoencoders compress heterogeneous patient histories into latent vectors

• Class imbalance handling: distribution-based feature augmentation to model ultra-rare diseases

• Predictive modelling: CatBoost, gradient boosting and deep neural classifiers for multi-class probability scoring and high-risk thresholds

• Explainable AI: SHAP charts show clinicians exactly which biomarkers drove each risk score; ICD-10/11 and OMIM codes in the evaluation view

• Integration: REST API and FHIR-compliant exchange with hospital information systems (HIS); one-click referral for confirmatory genetic and biochemical testing

• Platform stack: Docker, Kubernetes, Linux (CentOS/Ubuntu), PostgreSQL, Redis, MongoDB

Target Pathologies

Calibrated for screening of four rare metabolic and storage disorders: Mucopolysaccharidosis Type I (MPS I), Fabry disease, Pompe disease (GSD II) and Niemann-Pick disease Type A/B (ASMD).

Target Industry VerticalsGeneral medical and surgical hospitals, specialised paediatric hospitals, medical and diagnostic laboratories, diagnostic imaging centres, regional healthcare departments and digital health programmes, research and diagnostic medical centres, clinical research organisations.

Deep-Dive ReferenceExplore the full architectural whitepaper and engineering case studies in My Projects: https://www.loranjacobs.com/my-projects/neuroscanner-rare-diseases

Deployment & Licensing Matrix

• Delivery models: SaaS (hosted service); On-Premises (installed by our engineers inside the hospital intranet, including air-gapped networks, where data confidentiality requires it); DSaaS (Data Science as a Service: you provide high-quality raw data, we train and calibrate the models)

• Licence: standalone out-licensing with milestone-based payments; the same price applies to all delivery models

• Integration: EMR/EHR data lakes, HIS, FHIR, REST API

• Maturity: TRL 8 (system completed and qualified; successfully deployed in industry)

• Price: €3,600, excluding sales tax

• Initial deployment and support: from €72,000

• Average annual support and total cost of ownership: from €24,000 per year

• Final pricing is confirmed after approval of the technical specification (statement of work)

Clinical decision support tool. It supports, and does not replace, the judgement of a qualified clinician. Regulatory and clinical validation requirements depend on the jurisdiction and are confirmed during project specification.