In 30 years as an oncologist, Dr. Howard Fine estimates he has treated some 20,000 patients with glioblastomas, the most deadly form of brain cancer, "and almost all of them are dead." Of the 100 new glioblastoma patients he saw last month, "five years from now, only three will be alive," he said.

During a conversation recently in his office at Weill Cornell Medicine in New York City, Fine rattled off more dismal stats, like the many failed clinical trials of experimental drugs for glioblastoma; like the paltry increase in life expectancy for people with glioblastoma from 12 months in 1990 to 15 today; like the stupid (in hindsight) assumptions about how glioblastomas grow and how to study them in mice. Then Fine, 59, paused for several long seconds.

"My stance as an old man in this field is, someone has to start doing something different," he said. He thinks the "something different" just might be human micro-brains.

In the barely three years since biologists discovered how to create these "brain organoids," the lentil-sized structures have taken neuroscience by storm. Starting with a recipe developed by scientists in Austria, researchers from Japan and China to Europe and North America are seeding lab dishes with human stem cells, adding special molecules — many labs, like chili chefs, have their own secret blends — that make the stem cells morph into a variety of brain cells. They then put the dishes into special chambers called bioreactors that keep them warm and in gentle motion reminiscent of a womb, encouraging the cells to form blobs with working neurons and many other features of a full-size human brain.

Most of the researchers making mini-brains hope they will reveal what goes wrong in neurodevelopmental disorders such as autism and epilepsy, in mental illnesses such as schizophrenia, and in neurodegenerative diseases such as Alzheimer's. Fine is one of the few scientists using organoids to study brain cancer in hopes of personalizing glioblastoma care to an unprecedented degree: by screening drugs in mini versions of cancer patients' actual brains containing their actual tumor cells.

He had long despaired of oncology's dirty little open secret: that when scientists transplant bits of human glioblastomas into mouse brains, the standard research approach, the result doesn't mimic what happens in people. One problem is that the mouse brain is very different from the human brain. Another is that the transplanted bits of tumor act nothing like cancers in actual human brains, Fine and colleagues reported in 2006: Real-life glioblastomas grow and spread and resist treatment because they contain what are called tumor stem cells, but tumor stem cells don't grow well in the lab, so they don't get transplanted into those mouse brains.

Glioblastomas in lab dishes and mouse brains are fakes, little Potemkin villages that everyone thought were faithful replicas of human glioblastomas but which, lacking tumor stem cells, were nothing of the kind. In particular, the tumors put into mouse brains are nowhere near as invasive as patients' glioblastomas, which send out hundreds of invasive tumor cells and deadly little tendrils throughout the brain — and beyond the reach of surgeons.

No wonder what biologists learned from their glioblastoma cell cultures and glioblastoma mice was mostly irrelevant to glioblastomas in actual patients. No wonder experimental compounds that eradicated glioblastomas in mice failed in people.

"The mouse models don't recapitulate the human disease," said Ravi Basavappa of the National Institutes of Health, which gave Fine one of its 12 Pioneer Awards for "unusually bold," high-risk, and potentially high-impact research. "The hope is that [Fine's organoids] will be a promising path forward."

In 2014, Fine read the paper that launched the brain organoid revolution. It explained how biologists had coaxed human stem cells to develop into uncannily realistic miniatures of human brains.

"That's what we need," Fine recalls thinking. He called the authors and got more details about their organoid recipe. "We putzed around and tried different things," he said, and after some false starts — they weren't using quite the right brain-making ingredients, so the stem cells developed into micro-pancreases and colons — it worked.

"Within six weeks, we now have a brain organoid" with three of the six cortical layers that full-grown human brains have, he said. They even birthed specialized neurons that crackle with electrical activity and form circuits.

Now he and his team are putting cells from human brain tumors into the organoids, which have reached the level of development and complexity of a 20-week-old human fetus', to see whether they reprise what happens in patients.

According to his unpublished findings, when he puts glioblastoma cells from patients into lab dishes with brain organoids, the cells attach to the surface of the organoids, burrow into them, and within 24 to 48 hours grow into a mass that eventually "looks exactly like what happened in the patient's own brain," Fine said.

For one thing, the mini-tumors sprout microtubes that connect individual tumor cells and seem to underlie glioblastomas' resistance to chemotherapy and radiation. For another, the tumors in the brain organoids "mimic how far and how fast" the patient's own cancer grew, "and how destructive it was," Fine said. "The tumor stem cells kill the organoid in two weeks."

His team's first brain organoids were created from the cells of healthy people. But now they are making them from brain cells of glioblastoma patients. "We need to get the [tumor] stem cells to grow in an environment much more like a patient's brain," he said. What could be more "like" a patient's own brain than a miniature one grown from her own cells, and therefore containing DNA identical to hers?

If he can get funding and clear regulatory hurdles, Fine will grow hundreds of brain organoids from individual glioblastoma patients and give the organoids cancer. Then he will add experimental compounds or approved drugs, including those for seemingly unrelated conditions such as heart disease or alcoholism, and see whether he can rescue the otherwise doomed micro-brain.

Fine got federal approval last year to try such a drug screen on one patient whom he describes as "well-connected," creating an organoid from her cells and adding bits of her tumor to it in hopes of throwing drug after drug at it until one vanquished the organoid's cancer. Her glioblastoma was too advanced, however, and she died before the desperation experiment could identify any potentially effective drugs.

It might seem that because no existing drug cures glioblastoma, Fine's quest to find a compound that eradicates cancer in a brain organoid must be quixotic. Except for one thing. Even when an experimental drug fails in clinical trials, there is often a patient or two — maybe 1 in 50, he said — whose cancer stops growing or even vanishes, at least temporarily. That suggests there are super-rare "unicorn" patients whose unique glioblastoma and unique brain, with its characteristic immune system and genetics, might be vulnerable to a drug or a combination of drugs that, overall, doesn't work.

No glioblastoma patient could survive hundreds of Hail Mary attempts to find the drug or drugs, out of thousands of possibilities that might help her. But her brain organoids might.

This story was produced by STAT, a national publication covering health, medicine, and life science. Read more and sign up for their free morning newsletter here. You can also follow STAT on Twitter and like them on Facebook.