22 ms·

This reminds me of Paul Graham's recent list: http://news.ycombinator.com/item?id=3686840 http://news.ycombinator.com/item?id=3686840 Except there is no menti

by gsharm 15y ago

This reminds me of Paul Graham's recent list: http://news.ycombinator.com/item?id=3686840 http://news.ycombinator.com/item?id=3686840 Except there is no mention of cancer. Is a solution really so much out of reach? I feel like there's an identity issue here to blame, something along the lines of "we're programmers, we're not about making medical breakthroughs". Yet I can't shake off the notion that everything, absolutely everything, is interlinked, and we're in a better position, perhaps the only position, to understand and exploit this concept fully, in order to solve such problems.

dmix 15y ago
PG mentioned cancer 6 times in: 7. Ongoing Diagnosis This is a real problem hackers could solve: early detection of health problems which leads to preventative treatments. Although I lack any type of expertise outside of building a health startup I think this is probably one of the biggest ways we could reduce the impact of cancer on society... given the technology we have right now. It's not a matter of medical break-through though, but about building the right monitoring tools, creating meaning out of data, getting people to use them and the greatest challenge for startups: dealing with significant government regulation.
- bhickey 15y ago
  The sensitivity-specificity curve is going to be a roadblock. If your False Positive rate is 0.5%, True Positive rate is 100% and your incidence of disease is 0.1%, the probability that a positive result is indicative of disease is only 17%. Take a look at mammography, the false negative rate is estimated as high as 30% and the false positive rate of 65%. If you can stratify your population (ie. increase the disease incidence), you'll do a bit better, but it's an uphill battle. Sure, late detection is a factor in the median life expectancy with pancreatic cancer. However, I suspect that if tumors were detected 6 months earlier, it would move the median survival time by 6 months and make no dent in the 5-year survival rate.

by gsharm 15y ago

dmix 15y ago

PG mentioned cancer 6 times in: 7. Ongoing Diagnosis This is a real problem hackers could solve: early detection of health problems which leads to preventative treatments. Although I lack any type of expertise outside of building a health startup I think this is probably one of the biggest ways we could reduce the impact of cancer on society... given the technology we have right now. It's not a matter of medical break-through though, but about building the right monitoring tools, creating meaning out of data, getting people to use them and the greatest challenge for startups: dealing with significant government regulation.

bhickey 15y ago
The sensitivity-specificity curve is going to be a roadblock. If your False Positive rate is 0.5%, True Positive rate is 100% and your incidence of disease is 0.1%, the probability that a positive result is indicative of disease is only 17%. Take a look at mammography, the false negative rate is estimated as high as 30% and the false positive rate of 65%. If you can stratify your population (ie. increase the disease incidence), you'll do a bit better, but it's an uphill battle. Sure, late detection is a factor in the median life expectancy with pancreatic cancer. However, I suspect that if tumors were detected 6 months earlier, it would move the median survival time by 6 months and make no dent in the 5-year survival rate.