{"id":683,"date":"2014-04-11T00:33:53","date_gmt":"2014-04-11T00:33:53","guid":{"rendered":"https:\/\/linacolucci.com\/?p=683"},"modified":"2024-04-03T09:35:43","modified_gmt":"2024-04-03T13:35:43","slug":"medicine-diagnosis","status":"publish","type":"post","link":"https:\/\/linacolucci.com\/2014\/04\/medicine-diagnosis\/","title":{"rendered":"Medicine > Diagnosis"},"content":{"rendered":"
Health sensors will eventually diagnose every illness, disease, and ailment. In 20 years, diagnosis will no longer be part of the physician\u2019s job description.<\/p>\n
Doctors get agitated when they hear this. \u201cNever!\u201d they proclaim. \u201cMaybe for simple ailments but machines certainly can\u2019t diagnose complex, multi-faceted diseases. We\u2019ll always need talented physicians for those situations.\u201d<\/p>\n
I smile, nod politely, and think to myself how wrong they are.<\/p>\n
It worries me that doctors see their role as that of a diagnostician, when I see it as so much more. Notice I said that \u201cdiagnosis\u201d<\/i> will no longer be part of the physician\u2019s job description, not that a physician\u2019s job will cease to exist, which is what Vinod Khosla<\/a> believes.<\/p>\n Diagnosis involves the permutation of data until it fits into a clinical picture (disease model) with small enough error for us to believe that it is true. This diagnosis is either corroborated or shot down after we implement a treatment plan and see if data from the patient response is favorable or not. If it\u2019s the latter, the feedback loop of data continues.<\/p>\n Processing large quantities of information is exactly what computers were built to do. No physician can compete with the computational speed of a laptop. Computers don\u2019t forget and they don\u2019t have bad days. Computers are able to keep up with the deluge of new discoveries published daily in medical journals around the world.<\/p>\n Right now the main bottlenecks on the path to automated diagnosis are (1) that not enough data collection tools (sensors and devices) exist in a convenient form, and (2) that the data collection tools do not communicate with the computational tools (computers and smart phones) to make sense of those numbers. It\u2019s as if the speedometer on your car only<\/em> collected (and did not display) data, and you needed to use a special tool at the mechanic to print out your car\u2019s speed history. The speedometer wouldn’t be very useful and couldn’t inform your driving because feedback was so inconvenient and delayed.<\/p>\n These problems in data generation and data analysis are on their way to being addressed as thousands of start-ups work on various consumer health sensors and the government works on mandating interoperability rules.<\/span><\/p>\n