Ida Arvidsson: Applications of AI in Medical Image Analysis

Many medical examinations are done based on various types of images. By using machine learning to analyse these images, some assessments could be done automatically. This has benefits such as reducing the workload for medical doctors, decreasing variations in diagnoses and reducing waiting times for the patient as well as improving the performance. However, there are still few AI tools that are used in the clinic and more research is needed.

In this talk we will look into a few applications where deep learning can be used for assessment of medical images. We will discuss some of the common problems when working with medical data, such as limited amount and limited diversity of data as well as uncertainties in ground truth. We will have a look at the problem of automatically grading prostate cancer in biopsies, something that could be used to earlier determine which patients who eventually will need treatment. Another example we will look at is detection of breast cancer using a point-of-care ultrasound probe together with machine learning. This could be a feasible tool for diagnosis in low- and middle-income countries that today are lacking any solution for examination.

Ida Arvidsson is a postdoctoral researcher at the Division for Computer Vision and Machine Learning at the Centre for Mathematical Sciences, Lund University. She defended her Ph.D. in applied mathematics in 2021, titled “Applications of Deep Learning in Medical Image Analysis - Grading of Prostate Cancer and Detection of Coronary Artery Disease”. Her main research interest is deep learning applied to medical images. She is working with projects applied to histopathology (grading of prostate cancer), dermatology (classification of skin cancer) and radiology (classification of breast cancer, detection of coronary artery disease and prediction of Alzheimer’s disease).