An innovative automated medical imaging segmentation web platform. (2024)

On the other hand, developing a customized automated segmentation software requires several expensive resources such as AI expertise, expensive hardware and large amounts of labelled images.

This innovative segmentation platform overcomes all these limitations. The user can start by labelling a minimum of 6 examples and with this the system will automatically generate a customized automatic segmentation AI model for their specific problem. The system does not require any type of image pre-processing and users can use the images obtained directly from the source machines. The system can work with both DICOM and NIFTI file formats, including their compressed variants.

The user is not limited to use a single image type for their segmentation problem, if they have several imaging modalities they can employ them all as long as they are correctly co-registered beforehand (for example using both T1 and T2 in an MRI setting, or both a PET and CT images). Additionally the system is orientation-agnostic. The user can mix images acquired in different planes (sagittal, coronal, axial or any other, for instance cardiac-planes), the only requirement is that every image-segmentation pair follows the same orientation. This also applies for predicting new segmentations, the user can provide images in any plane without any additional work.

The new segmented images can then be assessed and manually rated to automatically refine the segmentation model, making the entire process dynamic and iterative, this allows to increase the model capabilities by selecting the best results without additional manual labelling required.

The company is looking for partners working in the development of novel imaging biomarkers interested in signing a commercial agreement with technical assistance. They can benefit from the use of a cloud-based system, suitable for any image modality that is commercialized as a SaaS service with a periodic license fee. The company is also open to collaborate in the framework of an R&D project that requires the characterization of the images from large cohorts of patients.

There are many segmentation tools available but all focus on specific imaging and target problems but none with the focus on fully automated segmentation for new problems. This is an understandable scenario since developing a single automatic AI segmentation model requires great number of resources. Due to the fast medical radiology development and new targets of study and characterization requirements, it is imperative to produce new segmentation models. This innovative platform is able to do it quickly and cheaply.