
Spectral Preprocessing for Clinical ML: Methods Compared
Spectral preprocessing determines model performance more than architecture choice. SNV, MSC, Savitzky-Golay, EMSC, and rubberband correction compared.

Connect Bruker, Thermo Fisher, and Agilent FTIR spectrometers to clinical workflows. SpectraDx handles instrument control, spectral processing, ML classification, and HL7/FHIR output — so FTIR results reach clinicians in minutes, not hours.
Overview
FTIR (Fourier-Transform Infrared) spectroscopy works by shining infrared light through a sample and measuring which wavelengths get absorbed. Different molecular bonds absorb at different frequencies, producing a unique chemical fingerprint for every substance — the same principle behind breathalyzers and airport explosive detectors, now applied to clinical diagnostics.
In clinical settings, FTIR enables rapid pathogen identification, tissue characterization for cancer screening, and unknown substance identification — all without reagents or consumables. A single measurement takes seconds, and the instrument requires minimal operator training compared to mass spectrometry or PCR.
The challenge isn't the spectroscopy — it's everything around it. FTIR instruments produce data files, not clinical results. Translating a spectrum into a diagnosis requires classification models, patient context, clinician-friendly reporting, and integration with hospital information systems.
Pathogen Identification
Rapid bacterial ID and antimicrobial resistance profiling from clinical isolates
Tissue Analysis
Intraoperative tissue characterization for surgical margin assessment and cancer screening
Drug Identification
Emergency department substance testing — identify unknown compounds in minutes
Biofluid Analysis
Cancer screening from blood serum using infrared spectral signatures
Instruments
SpectraDx provides modular instrument adapters. Each adapter handles vendor-specific protocols so your clinical workflow stays consistent.
Bruker
Full DDE/Named Pipe integration via brukeropus library. Automated measurement, background collection, and data export.
Bruker
IR-based microbial identification system. Integration via REST API for automated clinical microbiology workflows.
Thermo Fisher
OMNIC SDK integration for the Nicolet iS50 and iS20 series benchtop FTIR spectrometers.
Agilent
MicroLab integration for compact, diamond-ATR FTIR workflow automation.
PerkinElmer
Integration via PerkinElmer Spectrum SDK for portable FTIR applications.
The Problem
Instrument vendor software like OPUS (Bruker), OMNIC (Thermo Fisher), Resolutions Pro (Agilent) is designed for research, not clinical deployment.
Platform
Direct control of Bruker Alpha II via DDE/Named Pipes. Automated background collection, sample measurement, and spectral acquisition without manual software interaction.
Learn more →Parse OPUS .0 files, apply preprocessing pipelines — baseline correction, normalization, derivatives — and extract features for clinical ML models.
Learn more →Deploy trained classification models that convert preprocessed FTIR spectra into clinical decisions with calibrated confidence scores.
Learn more →Send classified results via HL7v2 ORU messages or FHIR R4 DiagnosticReport resources directly to the hospital LIS/EHR.
Learn more →Resources
In-depth technical articles for FTIR integration, clinical workflows, and regulatory compliance.
From Bruker OPUS files to clinical results in your EHR. One platform, zero middleware.