Data Science for Medical Imaging and the Business Behind it
Data science is an industry which has been growing rapidly in recent years. It is a field which has many connections to a vast number of other fields. In this blog post I would like to discuss how data science can be used to improve the healthcare industry. Specifically, I would like to discuss how Artificial Intelligence can be used for preliminary diagnosis and automated test diagnosis. This AI has many applications, but in this blog I will discuss how it is used for the early detection of illnesses. I will use examples from recent breast cancer research, but this AI can be applied to all different types of cancers and other illnesses as well.
Statistics and Process
In the United States, 300,000 people are diagnosed with breast cancer every year and 40,000 people die from it. The way that the process works for diagnosing patients with breast cancer is that people first go in for a mammogram. The mammogram will be able to tell the doctor whether or not the patient has high risk lesions. If there are high risk lesions, then the patient’s next step is to go see a doctor and get a biopsy. These biopsies are very expensive and are often unnecessary. 70% of women who get these biopsies do not need them because they actually have low risk lesions. This means that the false positive rate for mammograms is 70%. This is an issue for a lot of reasons. The biopsy is expensive, painful, leaves scares, takes up a lot of time, and many more issues.
Business Understanding
A typical biopsy costs on average around $500. And around 1.6 million people are getting biopsies every year. So if you take this number and subtract the 70% of false positives, you are left with hundreds of millions of dollars that are spent unnecessarily annually. This extra money could be spent in many ways. For example, it can be used for breast cancer research or it could be used by insurance companies to lower premiums.
False Negatives
Additionally, 10% of people who are told after the mammogram that they have low risk lesions actually have high risk lesions. These people should have received a biopsy, but were mistakenly told that they did not need one. This is an issue from both a health and financial perspective. People who received false negatives have more expensive surgeries in the future and have a lower chance of survival. This is another reason why it is beneficial to implement this AI technology. It will be used to eliminate both the false positives and the false negatives.
Stakeholders
The most obvious and important stakeholder in this scenario is the patient. They are the ones that receive the false negatives and are misdiagnosed. They are also the ones that are told they need this painful, expensive and unnecessary procedure when really they do not need it. The second stakeholder for this AI technology is the insurance company. They are the ones paying the unnecessary hundreds of millions of dollars annually on these useless biopsies.
Risks and Mitigating the Risk
- Implementation. Hospitals have no incentive to implement this technology into their system. They do not want this because radiologists and other doctors are going to be put out of their job and have no salary if this technology can do exactly what they can but better. In this case, the robot will be analyzing the scans instead of the doctors. Furthermore, it will lower the revenue of the hospital because they will not be doing these procedures anymore. One may to mitigate this risk would be to pitch this idea directly to the insurance companies because they are the ones that are paying for the procedures. This will cause the insurance companies to go to the hospital and say that they will not pay for this procedure unless they use this AI technology that differentiate between the high and low risk lesions. The technology will be able to say how necessary the biopsy will be.
- Codifying the results of the Radiologists. One thing that is certain is that the radiologists do a poor job at differentiating between the high and low risk lesions. And the only way to make sure that the AI technology works to its full potential is by feeding it reliable scans and not false information. One way to mitigate this risk to take mammograms from years ago where we know they are reliable. Taking mammograms where doctors have followed up with the patients and found out their diagnosis. To make this robot as factual and reliable as possible you need to only input correct information.
Conclusion
This AI technology is beneficial from a financial perspective by saving hundreds of millions of dollars every year, and it is also beneficial from a health perspective where false positives and false negatives are going to be eliminated. Catching cancer at an early stage is proven to have saved lives and using preliminary test diagnosis and automated test diagnosis is the first and most important step to doing that.
