By harnessing the power of AI, 91亚色 Associate Professor Ali Sadeghi-Naini aims to help cancer specialists analyze medical images more efficiently, giving them more time to focus on patient care.
In oncology clinics, diagnosing disease, monitoring treatment response and tracking tumour growth often begin with a meticulous review of medical images. For oncologists and radiologists, this can mean spending hours measuring tumours and comparing scans to understand how a patient's cancer has changed over time.
Much of that work is still done manually, even when clinicians are carrying out many of the same assessments across large numbers of cases. "There are many occasions where the time of oncologists and radiologists is spent reviewing cases that are very similar," says Sadeghi-Naini.
Spending significant time reviewing scans and performing routine assessments, he says, leaves less time for patients who require more urgent attention.
Sadeghi-Naini and his team at the QUANTIMB Lab are working to change that.
The 91亚色-based lab, led by Sadeghi-Naini, develops AI-powered imaging tools to support cancer care with a focus on helping clinicians process medical scans more efficiently.
鈥淭he idea is to automate things 鈥 to detect the location of the disease, quantify it and generate a report so the clinician can quickly review it,鈥 says Sadeghi-Naini, who teaches in the .
Rather than replacing clinical expertise, the goal is to develop AI tools that act as decision support systems, helping care providers assess scans and prioritize care. The tools are trained using large collections of medical scans, allowing them to identify patterns associated with disease progression and treatment response, which can then be flagged for clinical review.
One area of the lab's work focuses on monitoring brain tumours using MRI scans. For some patients whose cancer has spread to the brain, disease may appear as numerous small lesions scattered throughout the organ. In some cases, Sadeghi-Naini says, patients may have 10, 20 or 30 tumours that must be identified and monitored individually over time to determine whether treatment is working.
To help with that task, the lab has developed AI models that can analyze MRI scans, identifying and tracking small cancerous growths across a series of scans. The technology can automate much of that work and generate information to support treatment decisions, reducing the time clinicians spend manually locating and measuring each tumour over successive scans.
Another long-running project examines how tumours respond to treatment. For patients undergoing chemotherapy, one challenge is determining early on whether a tumour is responding to the drug. In some cases, this may only become clear after months of treatment, by which point valuable time and opportunities to adjust a patient's care may have been lost.
Using ultrasound data, Sadeghi-Naini and his team train AI models to look for biological changes that signal whether tumour cells are dying. By assisting in identifying those signals earlier than conventional monitoring approaches, the technology could give clinicians faster insight into whether a treatment is working. Detecting those changes earlier could help determine whether a different approach is needed.
鈥淚f a patient can be identified as not responding as early as possible, that spares months of ineffective chemotherapy and unnecessary side effects,鈥 says Sadeghi-Naini.
Although the projects focus on different clinical challenges, they are built around the same idea: using AI to extract useful information from medical images and help clinicians make decisions more quickly.
Across these projects, collaboration and human insight play a central role. The lab works closely with hospitals and clinicians, including researchers at Sunnybrook Health Sciences Centre, to access patient data, test new technologies and ensure the tools being developed address real-world clinical needs.
Feedback and guidance from clinical collaborators on the brain tumour monitoring technology have already been positive, says Sadeghi-Naini, who notes that some have expressed interest in using it as a support tool.
The lab is also designing interfaces that make the technology easier for practitioners to use, recognizing that ease of use is key to adoption in real-world settings.
Collaboration with those working directly in the field also helps accelerate the work. The ultrasound project, for example, has advanced to a clinical trial, a milestone that reflects growing interest among health care experts in evaluating the technology.
Although much of his work involves advanced technologies, Sadeghi-Naini says the motivation behind it is simple: helping patients receive the right care at the right time. Tools that monitor disease more efficiently, or determine sooner whether a treatment is working, could improve decision-making throughout the course of treatment and offer more time to focus on patient care.
"At the end of the day, this clinician time can be spent on somebody who actually urgently needs it," he says.