Deep brain stimulation - implanting electrodes deep in the brain - could be a new way to treat severe anorexia nervosa, a Canadian study suggests.
Scientists studied 16 people with severe anorexia and found the treatment helped reduce depression and anxiety and in some cases prompted weight gain.
Researchers say further, larger studies are needed before the therapy could be considered more widely.
The study appears in the journal Lancet Psychiatry.
The research involved women aged between 21 and 57 who had had anorexia for an average of 18 years and had tried all other available treatment.
The women were severely underweight and researchers say some were at a risk of dying early because of the condition.
At the beginning of the study, electrodes were placed in specific areas of their brains, thought to be linked to anorexia.
Within a few months, some patients felt symptoms of depression and anxiety had improved.
And 12 months later, a number of the patients had gained weight.
The average body mass index of the group increased from 13.8 to 17.3.
Researchers also looked at brain scans before and after a year of electrical stimulation and found persistent changes in the areas linked to anorexia.
Dr Nir Lipsman, a neurosurgeon at the Sunnybrook Health Sciences Center, told the BBC: "There are currently no effective treatments for people with long-standing anorexia nervosa - people who are often the sickest and most vulnerable of dying from the condition.
"Our work, which builds on earlier trials, is one of the first brain-based strategies that has been shown to help with chronic anorexia.
"And my hope is that through this research we are also validating the idea that anorexia is a brain-based illness, not a personality or lifestyle choice."
But he accepted the treatment did not suit everyone in the trial.
One patient had a seizure several months after electrode implantation, and two people asked for their electrodes to be removed during the trial.
Writing in the same journal, Dr Carrie McAdams, of the University of Texas Southwestern, said: "Further work to establish efficacy, safety and long-term outcomes in a larger cohort is needed."
Prof Rebecca Park, of the Royal College of Psychiatrists said: "While these results are encouraging, we must remember that deep brain stimulation for anorexia nervosa is a high risk, experimental treatment.
"In Oxford, we are running the sole registered UK trial of this kind.
"Central to our work is the development of an ethical standard that ensures vulnerable individuals are not inadvertently exploited by this treatment."