Traditionally, diagnosing a patient’s condition has been based heavily on the umbrella approach, which is based basically on the patient-reported symptoms, resulting in an umbrella diagnosis and treatment that frequently fails to achieve their intended effects due to individual variability. To exemplify that, just in Europe these failures are responsible for up to 200,000 people deaths each year. In addition, this classical approach is tremendously time consuming and expensive. Precision medicine aims to allow tailor prevention and treatment with an individual approach. Due to the complexity of human biology, individualized medicine requires taking into account aspects that go well beyond standard medical care. In fact, medical care only has a relative contribution of 10 to 20 percent to outcomes; health-related behaviors, socioeconomic factors, and environmental aspects account for the other 80 to 90 percent. The main role of precision medicine is identifying and explaining relationships among interventions and treatments on the one hand—and outcomes on the other— in order to provide the next-best medical action at the individual level. These can leave us for more accurate, earlier and more granular results. With that in mind, how to collect and work with that amount of relevant data needed in a quantum computing model system to help to optimize the medicine approach on day-a-day use?

What it does

The necessity for more personalized medicine, is currently very apparent as the stress on doctors increase, and the amount of information to consider about a disease have become numerous. This project aims to help with decreasing the burden on the health care system and providing more reliable care.

That’s where the QuDoc comes in. It is a precision based approach for health care platform, which with a user friendly interface can collect a large amount of relevant data from our patiences. Assessment of risk factors of cervical factors to produce a risk probability of developing the disease connected to a user app. Input data like recent symptoms, birth and gender, residence and travelling, lifestyle details, genetic exams, family tree, allergies information, and disease history. As the data about diseases increase due to the high rate of progression in the life sciences, we would be able to integrate all of the risk factors and produce a reliable probability of an individual developing a disease.

Diagnosing a patient correctly has always been a long and arduous, that depends on unreliable anecdotal data, such as patient-reported symptoms and vague familial data. This causes the misdiagnosis, or the prescription of ineffective medicine. This umbrella diagnosis and prescription is responsible for up to 200,000 people deaths each year. The use of personalized medicine seeks to resolve/attenuate this through consideration of individual variations such as differences in their genome, proteome, and exposome (takes into consideration their lifestyle, health history, etc..). This goes beyond standard medical due to the logistics involved, not only in collecting the data necessary but also in the processing of the data in a relevant way to prevent and treat the disease. To that effect, the role of precision medicine is to consider the patient at an individual level, using a multitude of factors, to produce an optimum solution that is tailored to them at that moment. This not only alleviates the repercussions of umbrella diagnoses, but also takes some of the burden off of healthcare workers who are experiencing burnout and declining mental health. The use of traditional machine learning has helped the progression of precision medicine but is coming to its limit as the number of health related variables increase. We can use quantum computing to aid this cause, and emphasize the interplay of health risk factors in the progression of disease.

How we built it

  • Python
  • Qiskit
  • Pennylane
  • PyTorch
  • Google Colab
  • Juypter Notebook

Challenges we ran into

  • The translation of classical data to quantum relevant information for our system to run with feature extraction.
  • Using Qiskit with PyTorch
  • Finding a medical data set with a large sample size and good features
  • Setting the hyperparameters for our quantum neural network
  • Time constraints and coding on Google Colab collaboratively
  • Code is currently on a local simulator. If we want to run quantum code on an actual IBM Quantum hardware - use unique API key locally.
  • Choose the features and matrix parameters for the quantum circuit

Accomplishments that we're proud of

  • The presentation and the foundation of our project.
  • As beginners, we learned a lot which was our main goal :)

What we learned

  • The difficulty and certain problem that is currently facing the health workforce.
  • Quantum Algorithms like CNN, qGAN, and optimizing probability distributions - specifically for precision medicine (We also looked at Travelling salesman problem, QAOA for optimization, and Grover's search algorithm.)
  • Technical mentorship from the quantum community
  • Using Qiskit with PyTorch

What's next for QuDoc: Quantum Machine Learning Powered Healthcare Platform

  • Recommend individual plan of action.
  • More precise treatments and best outcomes.
  • Add better classical data processing with data mining
  • Include health diagnosis and medicine recommendations.
  • Optimize efficiency of code.
  • Add privacy considerations for medical data.

Built With

+ 1 more
Share this project: