Years before Alzheimer's ravages memory and robs its victim of the ability to dress or read, scientists believe there are tell-tale signs. Bit by bit, researchers are closing in on a range of tests -- including brain imaging and spinal taps -- designed to reveal those signs and predict who will develop the mind-crippling disease.

''We know that Alzheimer's is probably going on in the brain for decades before people develop symptoms," said Dr. Susan Molchan, an Alzheimer's specialist at the National Institute on Aging. ''Eventually, we hope to be able to tell individual patients . . . what their risk is with greater certainty."

In September, the effort to develop new tests will get a major boost from the start of a five-year, $60 million project overseen by Molchan. The National Institute on Aging will provide about half the funds for the project -- designed to try out tests on hundreds of seniors -- with additional money from drug and diagnostic companies.

Work on treatments is proceeding on a parallel track, and researchers hope that the tests, when perfected, will also help determine who will benefit most from preventive therapies. Currently, there is no cure for the disease and only a few medications that alleviate symptoms.

About 4.5 million Americans now have Alzheimer's, but that is expected to jump nearly three-fold by 2050 as the population ages, unless a treatment is found that delays or stops the disease. Millions more people have mild cognitive impairment, a less severe problem with thinking and memory that may lead to Alzheimer's.

Genetic studies can only predict about 5 percent of cases; the rest of Alzheimer's victims are struck, seemingly, by chance.

Even diagnosing the disease is tricky. An experienced medical team can diagnose Alzheimer's with about 90 percent accuracy, using a battery of tests that includes all of the following: a physical examination with blood and urine tests, a brain scan using CT or MRI, a psychiatric evaluation, neurological tests to measure memory, language skills and reasoning, plus extensive interviews of the patient and family members. The only foolproof way to diagnose the disease is with an autopsy after a patient dies.

In addition to helping identify the disease early, scientists say they hope to use the new tests to determine which promising treatments work well, and to streamline diagnosis.

Researchers say the most promising tests include taking pictures of the brain with MRI and PET scanners and measuring levels of various compounds in spinal fluid. But they are looking at everything from proteins in the eye to the brain waves produced while listening to odd tones. Most scientists believe successful early diagnosis will depend on a mix of these tests.

''Within the next 10 years, I hope to be able to put someone in a scanner -- MRI (magnetic resonance imaging) or PET (positron emission tomography) -- when they're 65, and make a series of measurements and make a serious prediction of their odds of going on to Alzheimer's in the next three, five, and 10 years," said Dr. Clifford R. Jack Jr., a professor of diagnostic radiology at the Mayo Clinic in Rochester, Minn.

The scientific community is buzzing over work published earlier this year that appears to give doctors the ability to see the build-up of the beta-amyloid proteins in a patient's brain -- believed to be one of the destructive forces of Alzheimer's.

Researchers at the University of Pittsburgh led by Dr. William E. Klunk developed a substance that binds to beta-amyloid and lights up in specialized PET scans. In tests on 16 patients with Alzheimer's, the substance collected in the areas of the brain most affected by the disease, but it washed out of the brains of eight of nine healthy volunteers, ages 21 to 80.

Another group of researchers at the University of California at Los Angeles has developed a different chemical that lights up both beta-amyloid and tau tangles, another destructive hallmark of the disease.

While imaging amyloid is new, research has piled up showing that traditional PET scans can distinguish between patients with Alzheimer's and those with other dementias based on how the brain uses glucose, a sugar that fuels brain activity. Dr. Norman L. Foster, a neurology professor at the University of Michigan, says the scan highlights loss of nerve cell connections that may precede the nerve cell death that makes Alzheimer's so destructive. Nerve cells use less glucose as they deteriorate.

Foster's tests on people with mild cognitive impairment found those with Alzheimer's-like abnormalities on the PET scans were more than twice as likely to develop Alzheimer's as those without. Preliminary work by researchers at New York University found that decreases in brain metabolism in the entorhinal cortex, a part of the brain involved in memory, predicted development of mild cognitive impairment in healthy seniors.

But both Foster and Molchan caution that the PET scans are not ready for widespread use in early diagnosis because they yield too many false results and can cost up to $3,000 per scan. The use of PET scans got a boost last month, though, when the federal Medicare program said it was planning to pay for the scans for patients whose Alzheimer's diagnosis was in doubt after thorough clinical examination, and for some patients in clinical trials who have early symptoms.

MRIs, which are less expensive and more commonly used than PETs, are also showing promise in diagnosing Alzheimer's. MRIs show shrinkage of the brain that occurs naturally with age, but is more pronounced in Alzheimer's. The shrinkage is most apparent in the hippocampus, which is crucial for memory, and the entorhinal cortex. It also shows the ventricles getting larger.

A study published late last year by Henry Rusinek, an associate professor at New York University, found that shrinkage in the medial temporal lobe of the brain as shown on MRIs was 89 percent accurate in predicting which healthy people over 60 would show mental decline during the next six years.

Scientists are similarly optimistic about testing spinal fluid for signs of Alzheimer's. One large-scale test showed that increased amounts of ''phospho" tau could distinguish between Alzheimer's and healthy patients more than 80 percent of the time. The new federal project will also likely test spinal fluid for declining levels of a form of beta-amyloid and for rising levels of isoprostanes, which indicate the level of damage to fats and fat-like compounds in the body, according to Dr. Christopher Clark, an associate professor of neurology at the University of Pennsylvania.

The tests are not yet ready for widespread use because most are not refined enough to differentiate disease from normal individual variations. In addition, the tests pick up changes that result for reasons other than Alzheimer's.

Still, most researchers believe a new battery of tests will be ready within a decade, about the same time that more effective treatments are expected to be available.

''What we'd like is a dirt-cheap, simple test that always worked," said William Thies, vice president of medical and scientific affairs for the national Alzheimer's Association. ''We're beginning to see some nibbling around the edges. We'd really like to identify the disease before it's symptomatic and stop it from progressing."