PHOTOPLETHYSMOGRAPHY WAVEFORM ANALYSIS DETECTS INFLAMMATORY STATUS IN PATIENTS WITH INFLAMMATORY BOWEL DISEASE

Background: Early detection of inflammatory flares and corresponding treatment adjustment were shown to improve inflammatory bowel disease (IBD) patient care. Current episodic inflammatory biomarker testing allows limited monitoring and is associated with suboptimal compliance. We aimed to test the feasibility of monitoring patient inflammatory status by non-invasive photoplethysmography (PPG) waveform analysis.
Methods: Patients hospitalized due to IBD flares and flaring outpatients treated with IV biologics were continuously monitored using a finger-based PPG device. Flares were defined as Harvey Bradshaw index>5, or partial Mayo score> 3, or fecal calprotectin >250mcg/gr, or C-reactive protein (CRP)>1.5X the upper normal limit. We extracted and analyzed features from the raw PPG waveform to predict levels of CRP, erythrocyte sedimentation rate (ESR), and platelet counts during hospitalization. Ambulatory patients were assessed at each scheduled infusion. Models were developed using support vector machines (SVMs). The dataset was partitioned into a training set, comprising 75% of the data, and a validation set, comprising the remaining 25%. Principal component analysis (PCA) on the PPG waveform features was conducted to assess potential distinctions correlating with levels of inflammatory markers.
Results: Monitoring results were available from twenty-six patients (14 Crohn's disease, 12 ulcerative colitis). Analysis of all pooled PPG measurements revealed that the PPG waveform analysis enabled prediction of ESR results with R² 0.91, and root mean square error (RMSE) 5.6; of platelet count with R² 0.87 RMSE 59.1; and CRP with R² 0.69 RMSE 1.55. Analysis of individual patient data demonstrated that PPG features reflected different ESR values along the treatment course as depicted in a representative example in Figure 1. Dynamics of PPG measurements were demonstrated in all monitored patients as compared to partial detection of inflammatory dynamics by conventional biomarkers.
Conclusion: Analysis of PPG waveform features allowed continuous passive monitoring of the inflammatory status in patients with IBD and was sensitive to daily dynamics in a heterogenous patient population. This technology may allow feasible, sensitive and continuous patient inflammatory state monitoring and support optimized IBD patient care.

Figure 1
PPG signal features reflect the changes in ESR values across treatment days

PCA analysis of PPG signal features.
Each dot represents a single measurement. Colors represent different days. Dot diameter is proportional to ESR values.

<!--![endif]---->