San Diego ? A tiny experimental corkscrew threaded deep into the brain can pluck out deadly blood clots and stop a stroke in its tracks, potentially giving doctors an entirely new tool against this major killer.

Nearly 90 percent of all strokes result from clots that lodge in the brain’s arteries, cutting off circulation and starving brain cells of oxygen and nutrition. The goal of the new device is to extract these clots before they do permanent harm.

“It’s like pulling the cork out of a wine bottle,” said Dr. Sidney Starkman, co-director of the UCLA Stroke Center. “What you want is to get the blood flowing back to the brain.”

The device is the most advanced of several new stroke treatment tools discussed Thursday in San Diego at a meeting of the American Stroke Assn.

The only treatment now available for the emergency care of strokes is TPA, the clot-dissolving medicine also used to stop heart attacks. Only about 5 percent of stroke victims receive it, and among those, it helps only about one in eight.

TPA can be used only within the first three hours after a stroke. Most victims arrive at the hospital too late or have other conditions that could cause TPA to trigger disastrous bleeding in the brain.

Starkman directed testing of the corkscrew — called the Merci Retrieval System — on 109 patients. All had suffered severe strokes within the previous eight hours, and none could be given TPA.

The corkscrew successfully retrieved the clot in about half of them. Three-quarters of these patients survived, and 40 percent had few or no lasting disability. Among those in whom the clot could not be removed, half survived and just 6 percent had a good recovery.

Starkman said many of the patients were paralyzed on one side or could not talk when the procedure began.

“What’s really remarkable is we can see the problem and pull it out,” he said. “In some patients, the moment the clot came out, they could move again or talk normally. It was instantaneous.”

Starkman’s study was financed by the corkscrew’s maker, Concentric Medical of Mountain View, Calif. Based on this study, the company has asked the Food and Drug Administration for approval to sell the device. An FDA advisory committee will review it later this month.

The National Institutes of Health will sponsor a larger study of the device. Patients will be randomly assigned to get the corkscrew or routine care.

Dr. Larry Goldstein, head of the cerebrovascular center at Duke University, said patients who received the corkscrew appeared to do better than expected. However, he cautioned that the many patients whose clots do not show up on brain scans would not be eligible for the treatment.

“It won’t be a panacea,” he said.

Another drawback is that it can be done only by specially trained neuroradiologists. A UCLA survey found that 82 percent of the population lives within a three-hour drive of a hospital that could do the procedure.

Dr. Morgan S. Campbell of the Alabama Neurological Institute in Birmingham presented preliminary results of another approach intended to improve circulation to the brain after a stroke. The treatment involves inflating balloons in the abdominal aorta, forcing more blood to the head.

The initial safety testing showed improved cerebral blood flow in 12 of 16 patients. The device, called NeuroFlo, was developed of CoAxia of Maple Grove, Minn.