Preventing dataset shift from breaking machine-learning biomarkers

Machine learning brings the hope of finding new biomarkers extracted from cohorts with rich biomedical measurements. A good biomarker is one that gives reliable detection of the corresponding condition. However, biomarkers are often extracted from a cohort that differs from the target population. Such a mismatch, known as a dataset shift, can undermine the application of the biomarker to new individuals. Dataset shifts are frequent in biomedical research, e.g., because of recruitment biases. When a dataset shift occurs, standard machine-learning techniques do not suffice to extract and validate biomarkers. This article provides an overview of when and how dataset shifts break machine-learning–extracted biomarkers, as well as detection and correction strategies.

3D imaging using scanning diffractometry

Imaging of cells is a challenging problem as they do not appreciably change the intensity of the illuminating light. Interferometry-based methods to do this task suffer from high sensitivity to environmental vibrations. We introduce scanning diffractometry as a simple non-contact and vibration-immune methodology for quantitative phase imaging. Fresnel diffractometry by a phase step has led to several applications such as high-precision measurements of displacement. Additional scanning may lead to 3D imaging straightforwardly. We apply the technique to acquire 3D images of holographic grating, red blood cell, neuron, and sperm cell. Either visibility of the diffraction fringes or the positions of extrema may be used for phase change detection. The theoretical analysis through the Fresnel diffraction from one-dimensional phase step is presented and the experimental results are validated with digital holographic microscopy. The presented technique can be suggested to serve as a robust device for 3D phase imaging and biomedical measurements.

Biometric Signals Estimation Using Single Photon Camera and Deep Learning

The problem of performing remote biomedical measurements using just a video stream of a subject face is called remote photoplethysmography (rPPG). The aim of this work is to propose a novel method able to perform rPPG using single-photon avalanche diode (SPAD) cameras. These are extremely accurate cameras able to detect even a single photon and are already used in many other applications. Moreover, a novel method that mixes deep learning and traditional signal analysis is proposed in order to extract and study the pulse signal. Experimental results show that this system achieves accurate results in the estimation of biomedical information such as heart rate, respiration rate, and tachogram. Lastly, thanks to the adoption of the deep learning segmentation method and dependability checks, this method could be adopted in non-ideal working conditions—for example, in the presence of partial facial occlusions.

Superimposing interferogram method using a multi-slit array to enhance sensitivity and interference definition of spatial-phase-shift interferometers

We propose a superimposing interferogram method to enhance the sensitivity and interference definition of spatial-phase-shift interferometers. The proposed method uses a multi-slit array as a redesigned field stop for the spatial-phase-shift interferometer. The widths of the apertural and blocked areas correspond to half the wavelength of the fringe pattern on the imaging plane and determine the wavelength that improves the interference definition while eliminating those wavelengths that worsen it, thus improving the interference definition of the interferogram. Additionally, to improve the interferometer’s sensitivity, several apertural areas were aligned to increase the amount of incident light. We performed spectroscopic measurements and confirmed improvements in both the sensitivity and the interference definition using a prototype of a bean-sized mid-infrared spectroscopic imager based on a one-shot Fourier spectroscope and a prototype multi-slit array. Although mid-infrared light is absorbed well by water and its intensity is thus weak for detection using the array sensors of cameras, mid-infrared spectroscopic imagers that incorporate our proposed method can be used for biomedical measurements of samples containing water.