PetuumMed: Physician Assistant

Artificial intelligence (AI) today is overhyped. What most companies call AI is really just ML — programs that allow machines to help people by doing routine tasks or processing large amounts of data. Machine learning is helping us do what we want to do better, faster, and more accurately. The benefits of widespread ML adoption are hard to deny if you consider any industry with many repeatable processes and large collections of data like manufacturing, finance, or healthcare.

Although some industries lack the software infrastructure and technical manpower to use machine learning, with the right fit, ML capabilities would drastically improve productivity, service, costs, and delivery. However, because these industries lack a technological infrastructure, there is very limited familiarity with and confidence in anything that calls itself “AI”. In order to allow these industries to gain access to ML capabilities, any product will need to fit seamlessly into current systems without high infrastructure overhead or workforce expertise in ML.

At Petuum, we think the best way forward is to develop a platform that addresses the different needs of specific industries. It’s not surprising that with all the hype around “evil AI,” many people fear losing their jobs to automation, even folks in highly skilled jobs like doctors. Knowing this, we are developing technology that supports people, instead of replacing them. Our ML solutions consider the user and their needs and desires around a machine partner. In healthcare, the most important thing is conforming the machine to fit a doctor’s expectations and helping without hindering — this means being unintrusive, accurate, and deferential.

ML can solve many problems that take up doctors’ valuable time by doing everything from processing data in Electronic Health Records (EHR), to predicting illness recession rates and prioritizing care. Specialized programs can extract symptoms, diseases, drugs, and other key information from medical notes to aid clinical workflows, and provide a second pair of “eyes” to detect indicators and diagnose disease from medical images.

With a smart machine assistant powered by ML, we can reduce the time-consuming and tedious processes that take doctors away from critical work and their patients. Our product, Petuum Med, is built for this. We are developing a vertical platform with six key functionalities specific to healthcare:

  1. Diagnosis — by processing and analyzing EHRs and clinical notes, Petuum Med can suggest diagnoses backed by data;
  2. Imaging — by analyzing medical images, it can detect and identify abnormalities and determine the corresponding causes;
  3. Similarity measurements — by comparing patients’ cases to historical data on treatments and results of similar cases, it can suggest likely solutions;
  4. Predictions — by processing clinical data, Petuum Med can predict rates of mortality and recession;
  5. Insurance — by processing forms and checking regulations, it can support ICD code filing;
  6. Information distillation — by intaking patient reports and notes, it can highlight key information to help doctors make decisions for admission, medication, and etc.

From these core functionalities, doctors can customize around a basic set of configurations. As the industry becomes more and more comfortable with relying on smart machines, ML applications in healthcare will grow and those using the technology will be able to customize and specialize it even further to meet their needs and develop their own unique functionalities.

Interested in diving into the technicalities? Stay tuned for our upcoming posts.