Ground-truth preparation in medical project

Maximize the potential of your AI-based medical projects with accurate and reliable ground-truth data preparation.

We’ve created a process of data preparation for medical imaging projects in which we use machine learning algorithms. Unquestionably, proper, objective data used to train the model will ensure the expected results.

LET'S TALK

Ground-truth preparation in medical project for models training

Pipeline for GT preparation

We have organized the whole process: from creating teams of experts, through both data annotation and its accuracy verification. Up to be sure the appropriate result is reached.

High degree of objectivity

Pipeline for ground-truth preparation in medical project ensures high objectivity. Moreover, it eliminates the risk of incorrect assessment resulting from a doctor personal experience.

The icon depicts two figures high-fiving each other. The figures are on a grey background. The hands high-fiving each other are depicted in a pink square. The figures are framed in a white frame.

Teams of experienced radiologists

The annotation team includes the most experienced doctors. That ensures the analysis of particular study is correct. Additionally, this process reduce the interobserver variability.

The pipeline of ground-truth preparation in medical project

1

Training data

We use data provided by the client or obtained by us, and we also leverage resources from proven, trusted vendors to develop advanced machine learning based models.

2

Team of experts

If necessary, we involve not only experienced radiologists but also other specialists. It is most important to ensure the maximum degree of objective analysis of a given study.

3

An icon depicting four hands putting together a jigsaw puzzle. The icon is maintained in various colours of grey. In addition, one of the puzzles has a pink colour.

Prepare annotation

Our task is to ensure an efficient both flow of studies between doctors and to monitor the whole process. Obviously, the team of experts is in charge of preparing the annotations.

4

An icon depicted a symbol of medical algorithms based on deep learning techniques. There is a gray brain image with neural network and also magenta square on the icon.

Evaluation

A second expert team reviews annotations, accepting or requesting improvements. Each study's annotation is double-checked during preparation and evaluation.

5

A test set chosing

After finishing the annotation process and receiving the expert final approval for all studies, a test set is then carefully selected from studies that were not used in training.

6

Validating

The studies' annotations from the test set are passed to a second radiologist from the evaluation team for assessment. Importantly, an expert has not previously assessed a study.

7

The process result

In sum, the test study and its annotations are evaluated three times. Finally, only after the study had been accepted by radiologists could it be used for model validation.

Gound-truth preparation in medical project: ensure objectivity

The entire workflow, also the correctness of the annotations and the data annotation, was organized by Graylight Imaging professionals.

We engage experienced radiologists with the expertise necessary to analyze a given study properly to prepare the data.

Our process ensures a high level of objectivity in analyzing a study. Moreover, it eliminates the risk of incorrect assessment based on a doctor’s personal experience.

The studies used for model validation and its annotations are verified in three stages. This process guarantees accuracy and reliability.