AI software develops MRI-quality brain images from CT scans 

Recently developed software offers diagnostic assistance for radiologists and other experts involved in interpreting CT scans. It was designed using deep learning, a type of artificial intelligence (AI).

This innovation is software that has undergone training to translate interpretations from MRI images to CT images of identical brain cases.

 Context 

While computed tomography (CT) is a budget-friendly imaging technology available in healthcare settings, even in regions with limited access to other imaging techniques, it is commonly perceived as less adept than magnetic resonance imaging (MRI) at visualising intricate brain structural changes and variations in ventricular system flow. Consequently, specific imaging tasks need to be done at specialised units within larger hospitals equipped with state-of-the-art imaging technology.

Michael Schöll, a professor at Sahlgrenska Academy who led the work involved in the study, carried out in collaboration with researchers at Karolinska Institutet, the National University of Singapore, and Lund University. He said, 

According to researchers, this represents a clinically established application of AI-driven algorithms with the potential to become a rapid and dependable decision support tool, effectively lowering the occurrence of false negatives.

These researchers therefore anticipate that this solution could enhance diagnostic capabilities in primary care, consequently streamlining patient referrals to specialist care.

The software was developed using images from a total of 1,117 individuals, all of whom had undergone both CT and MRI scans. The primary focus of the current study was on healthy older adults and patients with various dementia types.

The research team is also exploring another application related to normal pressure hydrocephalus (NPH). NPH is a condition that primarily affects older individuals, where cerebrospinal fluid accumulates in the brain’s ventricular system, leading to neurological symptoms.

It affects approximately 2% of people over the age of 65. Given the complexities of diagnosis and the potential for misdiagnosis, many cases likely go unnoticed.

In the case of NPH, initial findings suggest that this method can be utilised for both diagnosis and monitoring treatment effects.

This software’s development has spanned several years, and it is currently in ongoing collaboration with medical facilities in Sweden, the UK, and the US, along with a company partner.

This collaboration will subsequently prove to be essential for gaining approval and implementing innovation in healthcare.