������ý

Chemist Chuan-Jian “CJ” Zhong never expected to get into the cancer detection field.

Twenty years ago, thanks to funding from the U.S. Department of Defense and the National Science Foundation, he and his team developed super-small sensors with nanoscale olfactory films for detecting molecules in the air. The plan was to install them in the cockpits of jet fighters to alert pilots about dangerous levels of noxious gases from burning fuel or other sources.

This effort soon led Zhong’s team to collaborate with researchers who have pattern recognition and medical expertise and to explore the sensors for potential noninvasive breath detection of diabetes. A few years later, Zhong read about other researchers using his nanofilms as an “artificial nose” to detect other health conditions like lung cancer by analyzing patients’ breath for signs of illness.

“We thought, ‘Hey, since we have been working this sensor for a long time now, we should also do that,’” he says — and that pivot has led to innovations and fruitful collaborations on and off campus.

Zhong, a SUNY distinguished professor of chemistry, is among the researchers advancing cancer detection at ������ý. From lasers sorting healthy tissue from tumors to artificial intelligence sifting through mountains of DNA data, investigations at ������ý could benefit millions of people worldwide.

In Zhong’s lab, tables are filled with an array of equipment for chemical screenings and other experiments. Among them sit a few early prototypes of the breath analysis machine — none of them with Wi-Fi internet capability, he points out. (Later versions will get this sorted.)

Every few weeks or so, one of Zhong’s doctoral students drives to the Cooper University Hospital in Camden, N.J., to pick up breath samples. The plastic bags filled with air come from patients who have cancer at various stages as well as some control samples. A partner hospital in China also sends its results online.

To use Zhong’s cancer-detection prototype, a patient blows into the device, sending the breath’s volatile compounds across the sensors. If the cancer biomarker is detected, it creates an electrical signal that can be read on a portable, wireless device.

In addition to the sensor, hardware and electronics, the system also requires a database of information that the system can read. For that, Zhong’s team collaborates with Professor Shuxia “Susan” Lu from ������ý’s School of Systems Science and Industrial Engineering, who is an expert in database pattern recognition. Along with building a bank of information through data acquisition using the wireless sensor, this part of the project involves using AI to analyze the data.

Zhong sees the short-term goals for this project as straightforward ones: “Number one, we’re going to need a major grant. Number two, we need more patient samples, so we’ll continue to work with doctors to collect them. The problem is, it’s not like everybody’s coming to the doctor at the same time — one today, two tomorrow, then four the next week. The samples are spread out over many days, which makes it very hard to collect them.”

Looking further out, he would like to expand the development of the wireless platform on two fronts. One involves placing it in doctors’ offices for routine collection of breath samples, and other is a further-miniaturized version of the sensor that can be read by a cellphone, so people could check their breath regularly for any questionable combination of chemicals or earlier signals of cancer.

Two students are working on selling the personal breath-checker idea to potential investors through the University’s Excellence in Entrepreneurship and Discovery (EXCEED) Program, which looks to advance innovative research through funding, personnel support and entrepreneurship training.

“Right now, cancer screening is a CT scan, a blood test or other methods that are time-consuming and sometimes even invasive,” Zhong says. “We want to take a breath sample and diagnose: cancer or no cancer. This gives a patient an early warning to go to a doctor to double-check.”

A blood test for malignancy

More than 1.5 million Americans are diagnosed each year with solitary pulmonary nodules (SPNs). These abnormalities in the lungs, often found during routine X-rays or CT scans, are isolated groups of cells up to 3 centimeters in size.

Many SPNs are benign, but figuring out which ones are malignant isn’t easy. One method is to scan patients again in three to six months so the nodules can be rechecked. If they’ve grown or changed, there’s a risk it may be a malignant lesion and cancer cells already are traveling through the bloodstream to other parts of the body.

Another method is to do tissue biopsies, but those can be painful and difficult, because the nodules are relatively tiny. Missing the target and taking surrounding healthy cells instead can lead to misdiagnosis.

When both cancer specialists and imaging doctors find it hard to tell if a nodule is harmless, the doctor will take a close look at everything about the patient, from age and smoking history to workplace factors and results from other tests. Then, the patient and doctor will decide together if starting treatment right away is the best path.

Yuan Wan, an associate professor of biomedical engineering at ������ý, is developing a faster, less painful way to diagnose malignant SPNs. In 2022, he received a five-year, $2.4 million grant through the National Institutes of Health’s prestigious MERIT (Method to Extend Research in Time) Award. The program supports experienced researchers as well as early-stage investigators such as Wan.

Wan’s project focuses on analyzing extracellular vesicles, which are small sacks of proteins, lipids and nucleic acids that cells secrete for intercellular communication. A patient would give blood, and the vesicles would be extracted from the plasma and enriched using specially designed microfluidic devices.

Wan aims to reduce detection time so that patients know within a week whether their SPNs should be removed.

He hopes the research leads to wider analysis of the vesicles for DNA mutations caused by cancer. “Doctors will be able to tell which drug is perfect for a patient and can effectively kill the cancer cells,” he says. “They also can use the information to see whether the patient’s cancer is still progressing.”

Wan’s research group is also trying to narrow the number of patients who would need this test, so labs are not overwhelmed. “We need to be precise in selecting those who truly require the EV test,” he says. “Imaging can reveal lung nodules in many patients, but if all of them were to get liquid biopsies, the turnaround time would become very long, increasing the financial burden on both patients and insurance companies.”

The University of Pennsylvania’s Medical Image Processing Group is working with Wan’s team to analyze 3,000 CT scans using artificial intelligence, scoring lung nodules based on their size and shape to figure out how likely they are to be cancerous. If a nodule’s score goes above a certain point, that patient would be recommended for the EV test.

Wan is seeking funding to purchase more CT image data, because the more samples they have, the better the AI does. When tested on a third-party public database with around 800 CT imaging datasets, the program achieved over 90% in both sensitivity and specificity for diagnosing cancerous lung nodules.

“The goal is to use this additional data to further train our program and improve its performance,” he says. “We want to combine our AI-based imaging diagnosis with the EV test to see if this diagnostic strategy is effective.”

Laser-guided surgeons

When you shine light on an object, the wavelengths that reflect back are almost but not quite the same. On the quantum scale, a tiny number of photons transfer energy to the material’s molecular chemical bond, very slightly changing their color.

Raman scattering — named after Nobel Prize–winning Indian physicist C.V. Raman — is not something that can be seen with the naked eye, but sensitive equipment can spot the shades of difference.

Fake “Frank” Lu, associate professor of biomedical engineering at ������ý, uses the phenomenon to distinguish cancer cells from normal cells, since each scatters light differently. He sees the technology as safer than CT scans and X-rays, which use a potentially more dangerous form of radiation.

“This is a very clean technology that drives the chemical bond vibrations,” Lu says. “You turn off the laser, and the vibration stops. Nothing happens except a little thermal energy deposit. There’s no break in chemical bonds. There’s no electron loss. The tissue and the proteins will recover after this very short period of excitation.”

Lu has concentrated his research on gliomas, which are among the deadliest kind of brain cancer and cause 80% of all malignant brain tumors. In 2020, he received a $433,000 grant from the NIH to develop his method of label-free stimulated Raman scattering imaging that uses the properties of lipid droplets. The microscopic organelles are essentially packets of fat and oils that play multiple metabolic functions in healthy cells. It has proven difficult to study them in living specimens.

“A cancer cell contains a lot of lipid droplets, and they have been largely ignored in traditional pathology,” Lu says. “Chemical fixation and histological staining usually remove the lipid droplets. We have a perfect technology to image these droplets in their fresh, native condition in live cells.”

To differentiate healthy brain cells from cancer cells, doctors currently have two choices. One is to put pathologists on standby during surgery so they can do an immediate analysis, an option that Lu calls “very demanding, very stressful” because it can take a half-hour or longer to prepare the tissue samples and offer guidance. Alternately, the surgeon can close the patient and wait for a comprehensive pathology report, which could show the tumor has not been fully removed and another procedure is required.

If Lu can perfect his technique, he sees two quicker and less expensive alternatives. Surgeons could collect a tissue sample, and a technician could examine it using a Raman scattering imaging machine right in the operating room. Maybe even better, OR staff could slide a laser probe through an endoscopy tube into the brain itself, like someone scanning a dark cave with a flashlight.

“More precise operations are important so surgeons are not touching the neuron bundle but still can cut out more of the tumor,” Lu says. “If we have a fresh-tissue pathology approach, the cost and the time for surgeries can be significantly reduced, and surgeons can have more success.”

Building the data infrastructure

What if we already have the answer to cancer detection (and treatment), but we just haven’t unlocked it yet? Professor of Empire Innovation Nancy Guo thinks a lot about this possibility, and artificial intelligence may hold the key.

During the past 10 years, it has become routine for American doctors to order DNA testing for patient tumors to determine if they have certain mutations and, if so, to help set the best course of treatment.

“The U.S. is the only country in the world with health insurance that covers a patient’s complete genome sequencing for advanced cancer patients,” says Guo, a faculty member in ������ý’s School of Computing. “If you can justify it to improve patient care, the insurance will pay for it — so it can be earlier in their treatment. Patients don’t have to wait until the cancer already has spread everywhere to take the test.”

However, only a small fraction of that DNA information is analyzed for those conclusions. There are about 25,000 genes in the human genome, and identifying which ones are the most important for curing diseases is a monumental task.

That’s where bioinformatics can help. “Machine learning and artificial intelligence — all of these techniques — have not been fully applied to analyze this kind of data,” Guo says. “The data is there at the hospital and everybody pays for it, but it’s underutilized. I think this is a perfect time, and there is a pressing need.”

Guo has led multidisciplinary research into AI with funding from the NIH, the National Science Foundation and corporate partners. Using the latest tools for genome analysis and drug treatment, she has leveraged technology and infrastructure for detecting and fighting cancer.

For instance, one genetic test she helped to develop for lung cancer patients — now under review by the Food and Drug Administration — can predict whether tumors will return or metastasize. She and her team started thinking about how the same test could be an early warning before an official diagnosis.

“We took the gene assay we developed and tested against published data, and we said, ‘Can this go earlier?’” she says. “Maybe a suspicious nodule was detected, and after a biopsy we compare the tissue with normal tissue and make a prediction whether the patient has lung cancer or not. It can even be earlier than that, before a nodule is detected. We are also developing biomarkers in liquid biopsies.”

In clinical cohorts, the test shows about 95% accuracy for lung and breast cancer, although more research is needed.

Advancing Guo’s vision for precision medicine requires investments of money and expertise, as well as full (anonymized) access to DNA testing data from healthcare providers. She believes the potential discoveries could revolutionize cancer detection and accelerate drug development to make personalized medications a reality.

“It won’t be easy, but at least it’s not 10 years ago or 15 years ago when you didn’t even have the data,” she says. “The data is there. We just need to build this infrastructure. It needs to be multidisciplinary with a lot of collaboration, so that is a challenge. If we all work together, though, we can achieve it and then beyond.”