Fredrik Langkilde: More thorough evaluation needed before AI can assist in prostate cancer diagnostics
Prostate cancer is the most common type of cancer among men in Sweden, but its severity varies widely. Some tumors grow slowly and may not require treatment for years, while others progress rapidly and need immediate intervention to ensure a cure. Fredrik Langkilde's research explores whether MRI and AI can reduce the number of men undergoing tissue biopsies while enabling faster and more accurate detection of aggressive tumors.
FREDRIK LANGKILDE
Dissertation defense: 14 March 2025 (click for details)
Doctoral Thesis: Detecting and Localizing Prostate Cancer with MRI
Research area: Radiology and Imaging
Sahlgrenska Academy, The Institute of Clinical Sciences
The evaluation of prostate cancer has traditionally relied on a blood test, prostate-specific antigen (PSA), followed by tissue biopsies if the PSA levels are elevated. Tissue samples are collected using ultrasound guidance, with needles inserted into the prostate either through the rectal wall or the perineum.
"Magnetic resonance imaging (MRI) is now being used before tissue biopsies to better select which patients actually need a biopsy. This is because MRI can often rule out clinically significant prostate cancer," explains Fredrik Langkilde, a radiology specialist at Sahlgrenska University Hospital and a doctoral student at the Institute of Clinical Sciences.
Advanced MRI models provide no clear advantages
Langkilde’s thesis focuses on the development and evaluation of models for analyzing MRI images of the prostate, with an emphasis on diffusion-weighted MRI *. The study also examines whether artificial intelligence (AI) can enhance image analysis.
"Our findings indicate that more advanced models for analyzing diffusion-weighted MRI data offer no clear advantages over the methods currently in use. However, in the future, these models might help differentiate between different types of prostate cancers."
AI’s potential in managing large volumes of MRI images
Artificial intelligence has significant potential to streamline the processing of the large number of MRI scans required for prostate cancer diagnostics.
"However, our study showed that AI identified more false positives compared to experienced radiologists, which could have implications for diagnostics. The effects of this have to be studied before the technology is to be introduced everywhere," Langkilde notes.
Most rewarding and challenging about this project
The project provided Langkilde with a deeper understanding of MRI technology, AI systems, and their applications. One particular challenge was planning and executing one of the studies conducted using an experimental MRI scanner at Medicinareberget. The challenge lay in that few similar projects had been carried out before, requiring meticulous planning—from sample collection to conducting the MRI examinations.
"The most enjoyable aspect was collaborating with individuals from various professions and specialties. The Urology Department played a crucial role, but hospital physicists and computer scientists also contributed. I believe and hope that this collaboration was valuable for everyone involved."
Text: Susanne Lj Westergren
Magnetic resonance imaging uses magnetic fields and radio frequencies to produce images of the body without harmful radiation. These images can include detailed visuals of internal organs, as well as measurements of processes within the body. One such process is the random movement of water molecules, known as diffusion. The imaging technique called diffusion-weighted MRI is already used for examining the prostate, and studies have shown it is the most important component of MRI for detecting clinically significant cancers.