According to medical researchers, sophisticated systems are critical for real-time diagnostic and disease monitoring devices for the detection of biomarkers.
A team of researchers at UC Santa Cruz has long been engaged in developing unmatched, highly sensitive tools called optofluidic chips to find biomarkers. This involved an effort to use ML to upgrade the systems by enhancing their ability to precisely classify biomarkers. The deep neural network thus developed classifies particle signals that are 99.8 percent accurate in real-time, on a system that is relatively inexpensive and portable for point-of-care applications.
The findings are published in a new paper in Scientific Reports.
Meanwhile, the signals received by sensors on displacing biomarker detectors to a point-of-care setting or into the field may not be of as high quality as received in a lab or in a controlled environment. This may be due to various factors, such as the need to use inexpensive chips to reduce costs or environmental factors such as humidity and temperature.
The need to address the challenge of a weak signal led the team of researchers to develop a deep neural network that can detect the source of a weak signal with precision. This involved training the neural network with familiar training signals, instructing it to recognize potential variations that were seen so that it can make out patterns and detect new signals with extremely high accuracy.
To begin with, a parallel cluster wavelet analysis approach designed in-house detects if a signal is present. This is followed by a neural network processing the potentially noisy or weak signal and identifying its source. The best part of the system is it works in real-time, hence users are able to receive results in an instant.
