Researchers at the Rady faculty of health sciences are innovating breast cancer detection, diagnostics and prognostics. Using artificial intelligence (AI) to analyze medical images and genomic datasets, members of the Hu Lab are making advancements in precision medicine for the fight against breast cancer.
Pingzhao Hu is an associate professor of biochemistry and medical genetics at the University of Manitoba and joint holder of the Manitoba Medical Services Foundation Allen Rouse basic science career development research award. Hu has been working on AI-based approaches to breast cancer diagnostics and prognostics for the past several years.
In medicine, treatments are not generally developed for single individuals, but rather for populations. Increasingly, groups of patients can be identifiedusing genomics, allowing for more specific treatment targeting.
When it comes to breast cancer, diagnosis and treatment programs are undergoing a paradigm shift away from a broad, one-size-fits-all approach to one that’s more personalized. The use of more sophisticated diagnostic techniques, like genomics, allows researchers to better characterize the tumour and results in better patient outcomes.
“I want to look at this question at the DNA level to see whether they have a genetic profile difference,” Hu said.
Hu’s research employs deep learning algorithms, a form of machine learning which trains itself on complex datasets and makes highly accurate predictions. This method requires less manual input than regular machine learning algorithms. The use of deep learning AI in diagnosing breast cancer would potentially result in increased speed and reduced costs, while also delivering precision medicine to the patient.
One of the projects Hu is working on aims to analyze over 20 years of MRI images collected from women in Manitoba. It is recommended that women over the age of 50 undergo mammograms every two years. This screening process has generated a huge quantity of MRI images for both healthy women and women with breast cancer.
Hu is currently in the process of determining whether AI can help radiologists to interpret these images and identify tissue abnormalities earlier.
“[W]e are trying to use this type of AI technology to […] help radiologists to detect the cancer,” said Hu.
Hu is hopeful that the use of AI will result in earlier detection of breast cancer by up to two years, which may improve the patient’s prognosis.
Qian Liu is a graduate student currently working on incorporating AI with breast cancer diagnosis. Liu’s PhD program is interdisciplinary, with three contributing departments: biochemistry and medical genetics, computer science and statistics.
Liu’s project combines computer science and statistical testing methods to analyze breast cancer patient outcomes. By applying AI algorithms to MRI images and genomic datasets, Liu can solve clinical questions like whether a reliable indicator for breast cancer can be found in MRI images. Since diagnoses and prognoses for breast cancer are still mainly based on the genetic profile of the patient, an AI approach to MRI would save time and costs.
In addition to breast cancer research, Liu and members of the Hu Lab are working in collaboration with clinicians and researchers to implement their AI algorithms to fight other diseases, including COVID-19.
Liu was recently granted a fellowship to develop a deep learning model and apply it to whole genome sequencing (WGS) data of 10,000 Canadian patients with COVID-19. This would allow researchers to create a scoring system that would predict disease severity for each patient. The funding was provided by the CANSSI Ontario STAGE HostSeq Fellowship, which was created for researchers interested in COVID-19’s genetic factors.
In this area, Liu is analyzing chest CT scans of patients with COVID-19 using AI models to generate risk scores associated with disease severity outcomes.
Due to bottlenecks in rapid PCR testing kits for COVID-19, there have been calls to use CT scans as rapid tests instead. When Liu learned of this, she saw a possibility to implement her research.
“That’s how we decided to transfer or use our knowledge in analysis […] to the COVID-19 images,” said Liu.
Liu’s development of AI algorithms to rapidly analyze these scans would reduce the need for doctors to interpret these images and increase the feasibility of the test.
As an undergraduate student, Liu majored in medical imaging, with no computer science or genetics in her background. She largely credits Hu for creating a supportive environment for learning, and the collaborative nature of the lab for her personal development.
In 2021, Hu was awarded a U of M Graduate Students’ Association teaching award as well as the Health Sciences Graduate Students’ Association’s Ed Kroeger mentorship award.
Since 2014, students and trainees at the Hu Lab have won 69 international and national awards, collected almost $700,000 in award funds, and have published more than 130 peer-reviewed articles.
To any prospective students interested in such a dynamic area of research, Hu had a few words of advice. A strong background in computer science or statistics and a familiarity with the life sciences will be highly beneficial, and a strong sense of personal initiative will carry you far.
Given the interdisciplinary nature of research in biostatistics, computer science and life sciences, researchers in the Hu Lab come from a variety of backgrounds. Due to the variation in skillsets, it is a highly collaborative space.
“This may include computer science, engineering, statistics and also the life sciences, such as genetics,” said Hu.