Machine learning increases the accuracy of detecting care diseases

What if people could detect cancer and other diseases with the same speed and ease of pregnancy test or blood glucose meter? Scientists from Carl R. Wese Institute for Genomic Biology are one step closer to this goal by integrating machine -based analysis at bio -viewing technologies at a care point.

The new method, called Loka-marka, was reported in the journal and improves availability biomarker Detection by eliminating the need for technical experts to conduct an image analysis.

Traditional medical diagnostic techniques require doctors to send blood or tissue samples to clinical laboratories, in which experts scientists carry out test procedures and data analysis.

Current technologies require patients to visit hospitals to obtain diagnostics that takes time. Many people also have barriers in which more visits may not be financially or spatially feasible. I think we can change something by developing more technology of the point of view available to people. “

Han Lee, the first author of the study and graduate of the Nanosensors research group

Diagnosis of care points is performed and brings results at the patient’s care, whether at home, doctor’s office, or anywhere in between. This allows for lower tests, easy to use and quick tests that can help in the next steps. Some examples already adopted in everyday life include urine pregnancy tests, Covid-19 antigen testing sets and blood glucose meters that allow People with diabetes React to declines and spikes in blood glucose levels throughout the day.

In the field of care, scientists study new ways of integration of this type of technology in patient care conditions, such as visits to specialists such as oncologists or oral surgeons. This would help to shorten the time and financial burden of patients, while improving real -time decisions for doctors.

“Doctors say that they would like something similar to a bacterial infection. They do a test for you, and then send you home with appropriate antibiotics, which will treat specific bacteria you have,” said Brian Cunningham (CGD leader), professor of electrical and computer engineering. “So why not do a similar thing in choosing the right anti -cancer drug or to determine whether the medicine you have taken for several weeks, begins to work or not.”

Earlier, the group reported a new bio -sensation method called the absorption microscope microscopy of a photonic resonator or trolley to detect molecular biomarker molecules in the body whose presence and levels indicate healthy or disease conditions. PRAM allows detection of individual biomarker molecules, including nucleic acids, antigensand antibodies; Instead, typical bio -lovage techniques detect a cumulative signal of hundreds to thousands of molecules.

Cunningham said: “Basically, what we do, shines with red LED light at the bottom of the sensor. Then at the top of the sensor the molecules land and detect when they have a small particle made of gold, which we call gold nanoparticles or AUNPS-tied to it.”

Paintings generated with a trolley depict a red background with small black spots sprinkled with it. But although these images themselves seem relatively simple, obtaining an exact number requires a trained eye that can decipher what things really correspond to biomarkers marked by AUNP.

“There are many types of artifacts, such as dust particles or nanoparticles. If you don’t have much experience, it is difficult to distinguish them,” Lee said. “The conventional counting algorithm we use requires adaptation of many parameters to get rid of these artifacts.”

To lead this process out of the laboratory and make it better for care environments, Lee proposed integration of machine learning with the process of image analysis.

“Han really developed interest in machine learning after taking part in classes here at the university to find out about it,” said Cunningham. “One day he came to me and said that he thought that he could make a machine learning algorithm, which is more precisely our black spots.”

Compared to other biocolation techniques, PRAM is well suited for enabling deep learning algorithms, because it generates microscopic images, not just detection of optical signals. But because these algorithms are as good as the data that trained them, Lee decided to imagine the same samples using electron microscopy both a wheelchair and scanning.

AUNPS, which are 1000 times smaller than human hair and show only as small black spots in the pram’s paintings, can be more clearly visualized on an electron microscope. During the intensive time during Lee Cross, he appealed to any place in the paintings of the Pram with images of the electron microscope to get very accurate data for the machine learning training set.

“Finding the right place to compare was actually very difficult, because it’s like finding a needle in the desert. One of the ways I developed was to create a reference point, such as a lighthouse at sea. From there we can find exactly the same place for registration,” Lee said.

A deep learning method was created, called location with contextual awareness, integrated with PRAM, enables high detection of molecular biomarkers without the need for sight and experience of technical expert in real time. After testing, the team stated that Loca-RAM exceeded conventional accuracy techniques, detecting lower biomarker levels and minimizing false positive and negative indicators.

“The whole journey of my doctor began because I wanted to make changes in the field of care,” Lee said. “I just want to do everything in my power to develop more advanced technologies that can affect the future.”

Publication “Physically well -established location of gold nanopartins with deep learning and quantification in photonic absorption microscopy for molecular diagnostics of digital resolution” https://doi.org/10.1016/j.bios 2025.11455 I He was supported by the National Institutes of Health, USDA Afri Nanotechnology Grant and the National Science Foundation.