Cancer Detection With Machine Learning

Improved, AI-assisted solution to aid in detecting cancer cells in medical images.

    SHARE

How AI models can contribute to cancer research

As the rates of AI adoption are globally growing across many industries, the medical sector is one of those that can benefit largely from the use of artificial intelligence. From operational use cases that can help stock hospitals, plan occupancy, schedule the medical practitioners' shifts to advanced diagnostic aid, cancer research, and drug discovery - machine learning algorithms are becoming more and more useful at hospitals and other medical & research facilities.

Among the plethora of use cases for healthcare, one that's exciting, and at the same time challenging, is AI-assisted diagnosis.

Team

  • 2

Duration

  • 3 months

Team role

  • Senior Machine Learning Developers

Industry

  • Medtech
  • Healthcare

Technology

  • Python
  • PyTorch
  • NumPy
  • MonAI
  • Matplotlib
  • Neptune.ai
  • OpenCV
  • TorchMetrics

Artificial intelligence in cancer diagnosis

Let's start with answering the question: How does AI "see" the disease? To do that, let's first briefly look at what AI means.

Simply speaking, AI systems are based on machine learning models that "learn" from data. Artificial intelligence refers to a system that is trained to identify patterns in data to make predictions or decisions. Modern AI solutions rely on various machine learning algorithms - and though the two terms (AI & ML) are not the exact same, they are often used interchangeably. Machine learning is a subset of AI that involves statistical models and algorithms that can be trained on data to predict desired results or perform tasks.

To have a machine learning model that is able to produce predictions - which in case of diagnostics is giving a positive or negative result in terms of cancer outcomes, we, naturally, need to train it with data first. For medical purposes, it needs to be factual data. The machine learning model can be 'taught' to spot anomalies and signs of dangerous symptoms in medical imaging like CT scans, X-rays, MRIs. Systems that aid diagnosis are based on neural networks that assess the medical images alongside other patient data to provide information that helps doctors reach a diagnosis and adjust cancer treatment.

Cancer detection with Machine Learning

Artificial intelligence is more and more commonly used as a diagnostic aid for various medical conditions, including many types of cancer: lung cancer, breast cancer, brain tumors, skin cancer. In these use cases, AI is used to help doctors find answers to questions that they normally answer on their own, basing their decisions on cancer imaging. They have to tell an 'innocent' lump from cancer, learn how fast it's growing and how far it has spread, and whether it's growing back after treatment. Researchers and doctors are now taking advantage of AI's accuracy and speed to find the right diagnosis, improve cancer care, and boost patient outcomes.

Complex tasks that rely on "a human making an interpretation of an image—say, a radiologist, a dermatologist, a pathologist —that's where we see enormous breakthroughs being made with deep learning," he said.

AI-powered diagnosis aid is not a new concept, though. Such systems have been supporting doctors in the interpretation of e.g. mammograms for over 20 years, but are now being developed and improved to improve cancer screening methods. Though there's a lot of hype around artificial intelligence and it's reported that many organizations fail at benefiting from AI/ML systems, in case of healthcare, there's actually a lot of potential that can be used, especially for computer vision that can be utilized in clinical practice for early detection of cancer in medical imaging of various types.

" height="">

Our solution

We have our part in the research to develop better solutions to aid detecting cancer cells in medical images.

When the Client, a cancer diagnostics company, approached us, they already had a number of AI projects that assist in medical image processing and interpretation. The company is also focused on continuous improvement of the analysis process and their machine learning models.

At the beginning of our cooperation, the client was in the research phase of a project and needed to explore the possibility of using AI for cancer diagnosis. There was a need to support their existing ML team with team augmentation to boost the research velocity and improve cancer detection.

Our tasks involved developing, evaluating and testing various deep learning models to find those that would best fit the use case and prove efficient, accurate, and precise. We also performed an in-depth analysis of performance in terms of memory and CPU optimization.

Every step of the project needed to be thoroughly documented since the result was to be used as a base in a medical device.