“The superpower of these AI systems is that they can look at all of these large amounts of data and hopefully surface the right information or the right predictions at the right time”
– The Harvard Gazette, “AI revolution in medicine”
Artificial intelligence (AI) has the potential to save lives. In the field of medicine machine learning powered models are beginning to look at the human body in the same way as physicians.
But in order to do this the models need to be trained with precisely labeled image datasets. It is vital that computer vision for medical AI is accurate, diagnoses depend on it.
Therefore medical annotation for training data must be pixel perfect. Keymakr is helping to support this revolution by providing this kind of data for researchers.
In contemporary medicine doctors have found themselves increasingly overwhelmed by vast amounts of data. MRI scans now provide many more times the images and level of detail compared to X-Rays.
Whilst more data is always useful, too much can overwhelm radiologists and lead to important details getting missed. AI will not replace doctors but it could potentially provide them with the support they need to make better decisions.
Creating labels for most images is relatively straightforward. It is a simple matter for most people to identify a car or a cat and label them accordingly. Some images however contain information that requires specialist knowledge that only a small number of individuals possess.
Medical imagery falls into this category. Medical annotation also has to be quality controlled and validated by experienced and knowledgeable experts. This too presents a challenge in terms of assembling the data.
Keymakr Provides Solutions
A client at the forefront of AI powered diagnostics approached Keymakr. They were looking for imagery data that could be used to train AI models to recognise brain tumours in MRI scans. Responding to this challenge Keymakr took the following steps:
- Provided the client with 2000 MRI scans of a head section containing a brain tumour, and 1500 MRI scans of a healthy brain without any tumour.
- All of the images were annotated by certified radiologists who specialized in tumour diagnosis.
- Their work was rigorously cross-checked by each other, and then verified by a senior radiologist in order to ensure that no errors slipped through the cracks.
- Polygon annotation tools were used in order to precisely identify the shape and location of each tumour.
- Strict standards of security were followed in order to keep the data and information gathered private and secure.
The data annotated by Keymakr was input into an AI medical computer vision model. The model was subsequently shown to be able to identify brain tumours with an extremely high rate of accuracy. Supported by accurately labeled data this model has the potential to catch abnormalities when they are still treatable and identify missed diagnoses.
Supporting Researchers With Professional Data Annotation
Medical annotation is helping to realise the potential of medical AI. Smart, professional annotation services can work collaboratively with developers on the cutting edge to craft bespoke image annotation solutions that enhance machine learning projects.
Keymakr utilises proprietary technology, in-house teams of annotators, and multiple layers of quality control to provide data annotation that is precise, affordable and scalable. Contact a team member to book your personalized demo today.