Image segmentation for more effective HIFU treatment
High-intensity focused ultrasound HIFU is a treatment for destroying tumours with heat. As part of the Intelligent Monitoring Programme of SalWe Ltd. a method for segmenting images that improves the therapeutic results of HIFU treatment has been developed. The work has been done by VTT Technical Research Centre, with Philips as the industrial partner.
A beam of high-intensity focused ultrasound (HIFU) is a non-invasive way of treating tumours by destroying them with heat. The therapy is currently being developed to make it suitable for treating the most common forms of cancer, in prostrate, cervix, bone, liver and breast. The advantage of ultrasound is that it can eliminate a tumour without surgery. It can be used alone or in combination with chemotherapy or radiation therapy.
The market potential of HIFU equipment is expected to reach 100 million dollars within five years.
Used to treat uterine fibroids
In HIFU treatment, a beam of ultrasound is focused on the affected area. The dose of heat delivered is precisely controlled by monitoring the area with Magnetic Resonance Imaging (MRI). At present this method of treatment is used for uterine fibroids, but its applicability for treating prostate and liver cancer is being studied.
“Image segmentation is an important tool in this method of treatment. It speeds up the planning phase and evaluation of the treatment’s effectiveness, as well as improving its precision,” says Kari Antila, Senior Scientist at VTT Technical Research Centre.
He has been responsible for the project to develop a method of segmenting images of the uterine fibroids being treated with high-intensity ultrasound. The industrial partner in the project has been Philips.
Philips in Helsinki, Finland is one of the forerunners in the research of HIFU technology and in the development of MRI guided HIFU treatment as well as in the commercialization of HIFU equipment.
Improvements at various stages
Image segmentation is valuable at several phases in HIFU treatment.
When the treatment is being planned, it identifies either the volume to be treated or the areas that should not be exposed to ultrasound, so that the configuration of the beam and the position of the patient can be set.
After treatment, the shape and volume of the tissue destroyed is displayed, so that a decision can be taken on whether therapy should be continued.
In the project, VTT and Philips have used image segmentation to display the outline of the uterine fibroids and estimate the non-perfused volume (NPV) after treatment.
Automating segmentation
“We have developed an automatic method for segmenting the image of the fibroid’s surface and the dead tissue. As input it takes the magnetic resonance image and data of the cell cluster being treated. It constructs an ad hoc surface around the cluster and uses a custom algorithm to deform it to the fibroid’s surface. Once the surface has been delineated, the area of dead tissue is segmented by clustering.”
The algorithm used is currently being validated.
“In the training phase, we have achieved an average accuracy of 0.84 (Dice coefficient) for the non-perfused volume, and we expect to improve it further with new data. A manuscript is being prepared for a scientific journal. We aim to test the same method for prostate segmentation.”
More information
Kari Antila
senior scientist
VTT
kari.antila@vtt.fi
+358 40 834 7509
Gösta Ehnholm
docent, scientific consultant
Philips
gosta.ehnholm@philips.fi
+358 50 368 4768
