Researchers have developed a series of AI-based systems that can interpret pathology images and identify the presence and absence of metastatic cancer. The AI systems could lead to new and improved diagnostic methods and treatment.
A group of researchers from Beth Israel Deaconess Medical Center (BIDMC) and Harvard Medical School in Boston have teamed up to develop new diagnostic methods based on artificial intelligence (AI). Humayun Irshad, PhD research fellow at Harvard Medical School and one of the lead authors on the research, says that their group is using all kinds of different computational methods to improve diagnostic techniques.
“We are developing robust and efficient computational methods to improve diagnostic and prognostic assessment of pathological samples,” Irshad says. “These computational methods include region of interest detection; cellular and subcellular object detection and segmentation; discriminative feature extraction; and building classification and prediction models. These methods are developed using image processing, computer vision, machine learning, and statistical algorithms. Altogether these methods will lead to new pathophysiological insights, and to improved diagnostics and prognostics for guiding patient diagnosis and treatment.”
|Don't miss the Minnesota Medtech Week conference and expo, September 21–22, 2016, in Minneapolis.|
The research was performed by the Beck Lab at Harvard Medical School, and the group recently put their new methods to the test at the annual meeting of the International Symposium of Biomedical Imaging. The group showcased the new AI-based systems by having them examine images of lymph nodes to decide whether or not they contained breast cancer. The group’s AI systems competed against private research companies and universities from around the world, and the results were good enough to receive first place honors in two separate categories.
The evaluation showed that their automated diagnostic method was able to produce accurate diagnostic results in 92% of the images examined, nearly matching the 96% success rate of a human pathologist. However, when the automated diagnostic method was combined with the efforts of a human pathologist, the results were rather astounding.
“When we combined our computational method with a human pathologist’s diagnoses, we increased the pathologist’s AUC (area under the receiver operating curve, or simply their success rate) to 99.5%,” Irshad said. “This represents an approximate 85 percent reduction in human error rate. These results demonstrate the power of using computational methods to produce significant improvements in the accuracy of pathological diagnoses.”
There’s little doubt that when it comes to merging AI technologies with medicine, there’s a laundry list of concerns. The rapid advancement of AI technologies has sparked a genuine question in the medtech industry: will AI be a savior or demon for medical technologies? An important question to consider when we’ve seen how far AI technologies have come already — and one that Irshad and his colleagues have considered when it comes to the possibility of misdiagnosing things like cancer.
“Misclassification of biopsies or surgical cores contributes to either overtreatment or undertreatment of tissues,” Irshad says. “Despite knowledge of many factors associated with carcinogenesis, pathologists’ ability to accurately diagnose and prognose remains limited and subjective. These computational methods permit the systematic assessment of single-cell to tissue-level morphological dynamics and relationships across a wide range of biological mechanisms, and provide new pathophysiological insights of tissue morphology and biomarker expressions from normal to invasive tissue.”
Irshad says that with recent advances in microscopy platforms and 3-D imaging, their group hopes to begin developing novel 3-D methods that can add to the analysis of their AI-based systems. The aim is to eventually demonstrate a 3-D image analysis that can be successfully applied to the construction of accurate diagnostic models that can provide new insights into three dimensional phenotypes associated with carcinogenesis. The result would eventually be a collection of AI-based systems that can be used in different ways to study cell biology and its complex mechanisms unlike ever before.
“The development of these novel methods in computational image analysis and advanced microscopy could improve the ability of pathologists to make highly accurate, reproducible, and predictive diagnosis of both benign and malignant tissues,” Irshad says. “They may also supplement conventional approaches in accurate cancer diagnosis, potentially prevent over and undertreatment, and guide clinical management.”
Kristopher Sturgis is a contributor to Qmed.
Like what you're reading? Subscribe to our daily e-newsletter.
[Artificial intelligence illustration courtesy of Victor Habbick at FreeDigitalPhotos.net]
- Considerations for Third-Party Reprocessing Of Single-Use Medical Devices - Webcast
- Drowning in Big Data: Extracting Medical Device Quality and Safety Insights - Webcast
- Quality with Confidence – What You Need to Know About Digital Microscopes for Medical Device Quality Processes - Webcast
- Risk Management for Medical Device Manufacturers - Webcast
- 3 Steps for Designing the Ideal Medical Device Packaging System - Webcast
- Reducing Device Cost with Innovative Medical Materials - Webcast