Modern mobile phone imaging sensors wide availability and high quality have enabled development of low-cost imaging and sensing approaches that utilize the camera, including those which detect diffuse optical interactions and produce quantitative transcutaneous measures analogous to clinical techniques for bilirubin and oxygenation sensing. Concurrently, recent clinical studies report overestimation bias from dark skinned patients in transcutaneous bilirubinometery (TcB) and pulse oximetry. Here, Monte Carlo simulations of TcB and oximetery were used to investigate the source of possible racial biases in clinical measurements. Simulations of device calibration studies with dark, mixed, and light skinned cohorts were tested against groups with similar and different racial distributions. Results implicate a combination of tissue optics and biased enrollment in calibration studies for systematic overestimation in both TcB and oximetry. Next, identical Monte Carlo simulations were performed with a 2D image sensor capable of detecting spatially resolved diffuse reflectance. Quantification models were developed from simulated calibration studies where reflectance was extracted from 1 to 5 unique sensor regions of interest (ROI), followed by evaluation against test cohorts with different racial distributions. The results indicated overestimation bias in darkly pigmented subjects could be reduced through incorporation of an increasing number of sensor ROI’s. Models for quantification of bilirubin were then developed using clinical data from our mobile phone based TcB study, and increasing number of sensor ROI’s improved model performance (r2) . These results suggest promise for the development of mobile image-sensor based spatially resolved diffuse reflectance for improving accuracy and reducing racial bias in transcutaneous measurements.
|