SenoCAD Research GmbH is dedicated to:
- Designing innovative Artificial Intelligence (AI) powered methods that provide reliable and cost-efficient solutions to challenging clinical problems.
- Offering our healthcare applications as compliant cloud services, which could be used in the frame of multicentric clinical studies and, later on, in clinical routine.
It is our endeavor to:
- Improve the radiologist workflow thanks to state-of-the-art artificial intelligence methods and computer vision algorithms.
- Empower the radiologist by providing him predictive biomarkers and a therapy response score that cannot be obtained by simple visual inspection of medical imaging data.
- Provide semi-automatic AI-powered methods that leave the full responsibility of clinical decisions to the radiologist or oncologist.
- Design reliable and cost-effective computer vision & AI-powered methods that could be applicable in clinical routine.
We have designed innovative segmentation/classification algorithms for the following clinical problems:
- Fast semi-automatic 3D segmentation of tumors in MRI and PET images: The aim is to estimate the tumor volume, which could replace the largest diameter to assess the response to NACT. Medical image segmentation is a difficult clinical problem that will never be perfectly solved by computers, according to radiology experts. Hence, there is a real need for a fast-interactive delineation method whose aim is, if necessary, to refine a 3D tumor segmentation generated by the computer. To this aim, we propose a semi-automatic 3D segmentation algorithm that integrates state-of-the-art computer vision and AI algorithms. The proof of concept, performed with head-neck PET images, has shown the possibility to significantly improve the workflow.
- AI-powered radiomics: We are using non-supervised AI methods that are not data hungry to allow the discovering of predictive radiomic features, which can be used to perform the early prediction of the resistance to NACT.
- AI-powered early prediction of the resistance/response to NACT: The response prediction is a challenging research topic. Machine learning has been used to reliably perform the prediction of the response to NACT. The proof of concept has given promising results in the case of locally advanced breast cancer. Now, these results must be validated by a large-scale multicentric clinical validation.
Later on, our AI-powered method for the early response prediction to NACT will be applied in cancer of other organs. Other challenging problems will also be addressed, such as the triage of normal medical images.
For each of envisaged clinical problems; two phases are considered:
- Algorithmic Research & Proof-of-Concept: the aim is to design reliable and cost-efficient AI-powered prediction methods. The realization of the proof-of-concept is using either publicly available medical data, and/or data provided by medical partners participating in a collaborative research project.
- Clinical Research: the objective is to demonstrate the reproducibility of results obtained by the proof-of-concept, in the frame of a multicentric clinical study led by a consortium of medical/industrial partners. In this context, we propose our support for data anonymization, data storage, 3D segmentation, radiomics, training of the AI-powered prediction model, and response prediction provided as cloud services. Then, the results of the clinical study will be submitted for publication in a peer-reviewed journal.
Our reliable AI-powered healthcare applications will be commercialized as compliant and cost-efficient cloud services that could be used firstly for research, in the frame of clinical studies, and later on, in clinical routine.
Please, contact us using our Contact form, if you are interested to participate in a collaborative research project.