MammoScreen™ is anArtificial Intelligence (AI) solution that supports radiologistsin reading screening mammograms
Our Technology
The true value of our technology isn’t measured by how powerful it is, but by how much it empowers you.Winner of the DREAM Mammography Challenge, the largest competition ever organised in the field of Artificial Intelligence (AI) for breast cancer screening to date. We use the latest machine learning approaches to analyse screening mammograms.
Resting on cases interpreted by expert radiologists and using state of the art deep learning techniques, our artificial neural networks learn to detect and characterise suspicious findings.
MammoScreen uses the most up-to-date AI methods to take advantage of the information contained at lesion, breast and mammogram levels.
Explainability
Our AI is coherent across levels of interpretation and always ties the analysis of malignancy at the mammogram level to the specific lesion(s). It is not a black box. In order to do this, we ally a top-down approach which grants us an overall assessment for the mammogram to a bottom-up approach that analyses specific portions of it.
Reliability
Our AI software carefully brings to service various families of machine learning algorithms to perform a wide range of tasks: detection of potential lesions, characterisation of the malignancy of these potential lesions and pairing lesions from multiple views.
The algorithms are tested on independent sources neither seen nor used by the algorithms during the development process. This renders null the risk of biased performances and gives an exemplar of conditions encountered in real-life cases.
Resilience
Diminishing the variability of predictions and extending AI robustness are key success factors for us. Our solution exploits state-of-the-art preprocessing and machine learning approaches to cut down on possible biases between the data used for training and the data used in production.
System Integration
Installation
MammoScreen™ is easy to integrate in existing workflows, allowing a fully-remote installation.
Mammograms are automatically pushed to a virtual local server installed on site and then are sent to our cloud where the AI algorithms analyse the images. Once generated, the AI results are sent back and stored in your practice’s network. All the data transiting to and from our cloud are de-identified and deleted once processed.
Flexibility
MammoScreen™ is easily introduced into common radiology workflows. Most of the time, the software can be synchronised to your existing viewer work list or to your reporting software.
Reading Environment Neutral
MammoScreen™ can be deployed in any environment, from a single general practice to a large and fragmented health system. It can be used with any PACS and RIS.
Focus on Security
We use a secure cloud. We methodically and consistently de-identify the DICOM images before sending them out to the cloud and delete them from the cloud once processed.