The Knowledge Inference in Medical Image Analysis Lab conducts research to address the challenges of using machine learning algorithms when analyzing medical images.
Overview
The word "kimia" is written in Persian.
Medical imaging is a fundamental aspect of modern medicine. Images from various modalities, including MRI, CT and positron emission tomography, are generated and used in many clinical settings, such as for analyzing tumors and planning cancer treatment. In North America alone, healthcare professionals order hundreds of billions of images annually. Some procedures generate thousands of images that need to be analyzed.
The Knowledge Inference in Medical Image Analysis Laboratory led by Hamid R. Tizhoosh, Ph.D., conducts research at the forefront of mass image data in medical archives using machine learning schemes. The goal is to extract information that not only can support faster and more-accurate diagnosis and treatment of many diseases but also, and more significantly, establish new quality assurance based on the mining of collective knowledge and wisdom.
The lab's acronym — KIMIA — happens to sound like the Greek term "χυμεία," which means the art of alloying metals, also known as alchemy. Once upon a time, scientists assumed that they could find techniques to turn any metal into gold. Perhaps the gold of the 21st century is buried underneath or among big image data?
Research focuses
Research focuses include:
- Artificial intelligence (AI).
- Big image data.
- Multimodal search and retrieval.
- Medical image analysis.
- Design and training of artificial neural networks.
Exploring AI in medicine
The lab explores the applications of AI in medicine, particularly in medical image analysis and cross-relations to other patient data, such as molecular, lab and textual data. To overcome the challenges of dealing with big image data in pathology and radiology, the lab uses:
- Deep networks.
- Supervised, self-supervised and unsupervised clustering methods.
- Dimensionality reduction and generative models.
- Large language models.
- Cutting-edge computer vision algorithms.
Healthcare professionals can infer invaluable diagnostic and prognostic knowledge in a precise and timely manner by identifying patients diagnosed and treated similarly. The goal is to improve patient care by automatically generating consensus reports for healthcare professionals. This is achieved by searching for and retrieving relevant evidence, thereby reducing both intraobserver and interobserver variability in clinical decision-making.
Delivering personalized medicine
Delivering computational consensus to healthcare professionals by searching and matching multimodal data of previous patients can make personalized medicine more accurate in terms of diagnosis and treatment. Healthcare professionals can diagnose and treat patients with higher levels of confidence, resulting in fewer side effects and shorter treatment times. Also, patients may wait less time to receive diagnostic decisions from healthcare professionals, reducing or eliminating stress that biopsies cause.
Addressing tomorrow's challenges
With continued investments and improvements in imaging equipment, the availability and use of imaging continues to increase. But bottlenecks will occur due to the limited capacity of healthcare professionals to process and analyze these images. This pressure is likely to create additional strain on the quality of the analysis. This could potentially affect patient care and outcomes and ultimately increase healthcare costs.
Advances and developments in medical image analysis algorithms, AI models and machine learning will be critical in reducing the effects of these imaging-related issues. The Knowledge Inference in Medical Image Analysis Lab conducts research to address the challenges of using a multitude of computer vision methods, machine learning algorithms and AI models to analyze medical images.