How easy is it to detect cancer? In some cases, you might catch it in the early stages. If you go in for a mammogram or “check your oranges”, as one Delhi ad prescribed. Maybe, breast cancer could be detected early. But, unfortunately, the reality seems to be that many cancers are detected at advanced stages, like pancreatic or ovarian or liver or others. Could it be that there aren’t effective early detection methods or screenings? The issue with late detection is that there could be higher chances of sadder outcomes or even limited treatment options. While cancer may be deadly, it might be underfunded. That’s unfortunate.
In this golden age of AI, where people are relinquishing their creativity to allow LLMs to do their thinking for them to finish school or college assignments in seconds, can AI step in on a grander scale to solve a much more noble mission: help detect cancer? Maybe, we’re not at the “Elysium” stage in the Matthew Damon world, where technology could eliminate cancer altogether, but if this works, it could be a step in the right direction.
So, what could have told Wade Wilson or Walter White or William Butcher or Hazel Grace they had cancer at an earlier stage? It’s said that new blood tests are emerging in a New York facility that use AI to spot signs of cancer in super early stages. And this could be used to ascertain whether someone might have a potentially lethal infection, like pneumonia, as well.
There’s something called nanotubes, which are like really tiny tubes of carbon roughly 50,000x thinner than a strand of human hair and which glow, when exposed to certain conditions. Scientists have been tweaking these nanotubes to see how they react to different substances like blood and this can help figure out what’s in the blood, like a sign of disease. And it all depends on how the nanotubes glow. That’s why they say “keep glowing, sparkle more”.
But, the issue has been about how to interpret the data, because it’s said that the patterns are too subtle for human beings to make a determination. It’s like having a fingerprint sample, but not knowing how to identify it or what to cross-reference it with. That’s where AI could do the job. The data is loaded with a classification of which samples came from patients with ovarian cancer and those patients who don’t have it. Unfortunately, ovarian cancer is rare, but deadly, and is spread across hospitals, so getting data shared with researchers can be tough.
And yet, an algorithm was trained on available data from about 100 patients. The result? It’s said that the AI model was able to get better accuracy than the best available cancer biomarkers. On the first try. Mostly, nothing comes right or works on the first try, so this is a big deal.
Years ago, we envisioned AI as Schwarzenegger obliterating people. Now, it might be a saviour against cancer. Who’d have thought?
So, could tech triage all diseases? It seems like traditional diagnostics takes time, a luxury that critically-ill patients don’t really have. Where’s Dr Gregory House to diagnose a patient in 45 minutes when you need him?
Now, while the AI model was able to work on data from roughly 100 patients, it’s been described as a “Hail Mary pass”. Because for AI models to be effective, there needs to be substantial data availability. That’s probably why AI companies, like OpenAI, are facing copyright lawsuits, because maybe, they might not be as effective without widespread data. Medical data is said to be siloed within institutions and inaccessible to researchers. In India, could something, like the DPDP Act, make that availability harder? Would there come a time when access to private medical data could have an “opt-in” option?
You may have often heard about how healthcare infrastructure is inadequate or underfunded in India. Could this tangibly change things? Based on this research and methodology, could there be healthcare startups or companies that could create such a kind of diagnostic tool that wouldn’t require too much physical infrastructure?
That would mean that for this to work globally, nanotubes would have to be shipped across the world. Could India be that key nanotube hub? Based on the disease profiles of Indians, can nanotubes create early detection for India? While still expensive on its own, maybe MedTech startups in India could create solutions that are cheaper than the cost it would take to formulate in developed countries. Though, developing AI-powered MedTech solutions wouldn’t be cheap, it might require a good chunk of funding. Of course, getting healthcare data in India could be hard, if many regions lack electronic health records or digitized patient histories.
And the big thing to keep in mind is that even AI-generated diagnostics could be wrong. If certain actions or treatments are taken to treat a cancer that doesn’t exist, could that open a Pandora’s Box of complications? A pregnancy test could have a false positive or a false negative, maybe, so could AI-powered diagnostics.
Will AI be your bloody brother? Or are we injecting too much hope and being starry-eyed about the “transformative potential of technology”?
