E!113351 An MRI-based technology for early assessment of antidepressant efficacy in depression.

Worldwide, major depressive disorder (MDD) affects >300M people, annually causes suicide in 800k people and costs society ?1 trillion each year. Many antidepressants are available. Yet, finding the right drug for individual patients remains challenging, as 50% of the patients have not found an effective drug treatment after 1 year. Lately, research including MR imaging has shown promising results. We would like to extend this further, using advanced MRI image analysis including deep learning to create a software tool that can be used in assessment of patients with major depression. With our assessment tool, we aim to reduce the time to selection of an antidepressant that the patient responds to by 75%. This will have major health and economic benefits. This collaboration project originally consisted of four partners, each with different roles. NordicNeuroLab AS is responsible for developing the final market-ready product for evaluating the effect of antidepressant treatment, Amsterdam Medical Center is responsible for collection of MR data of patients with depression, Oslo University Hospital for developing a radiomics post-processing pipeline, while Vuno Inc. from Korea was responsible for developing deep-learning methods using MR data from Amsterdam. Unfortunately, Vuno Inc. decided to withdraw from the project and the consortium in July 2021 since they were not able to achieve their objectives due to the difficulties related to data-sharing (GDPR). The consortium explored various alternatives to enable GDPR-compliant data sharing between European partners and Korea, but without success. In the time after Vuno left the consortium, the remaining partners re-organized their responsibilities to continue delivering on project milestones. The validation of deep-learning methods and the development of the prediction model has been re-assigned to all the remaining partners, and we are encouraged by the initial results obtained using deep-learning on test data. NordicNeuroLab has continued the work on the cloud-based platform (nordicMEDiVA). We have onboarded consultants to help us streamline the deployment process at customer sites, and regulatory consultants to guide and assist us with regulatory clearance in the USA and Europe. We also employed part-time consultants to develop Diffusion Tractography and fMRI analysis modules which will be part of the end-product. Finally, we have focused on the development of a web-based viewer which will allow viewing of and interaction with image data, quality control, and reporting of the results. The viewer is built using state-of-the-art tools that allow multi-site collaboration and teleradiology. AUMC can show promising results that may result in a commercial product after some further research. We have shown, together with OUS via other projects, that nordicMEDiVA is suitable for the implementation of solutions like DEPREDICT. nordicMEDiVA is being continuously developed as a commercial product.

The prediction model has shown promising results, and the final anticipated impacts as stated above, are unchanged. The project partner, AUMC, is exploring ways of further commercializing the prediction model, including the collection of additional training and validation data to improve the algorithm further.

Worldwide, major depressive disorder (MDD) affects >300M people, annually causes suicide in 800k people and costs society €1 trillion each year. Many antidepressants are available. Yet, finding the right drug for individual patients remains challenging, as 50% of the patients have not found an effective drug treatment after 1 year. Lately, research including MR imaging has shown promising results. We would like to extend this further, using advanced MRI image analysis including deep learning to create a software tool that can be used in assessment of patients with major depression. With our assessment tool, we aim to reduce the time to selection of an antidepressant that the patient responds to by 75%. This will have major health and economic benefits. The project includes four partners from Norway, The Netherlands and Korea, both universities and small business entities. The project is divided into six work packages: MRI data collection, radiomics analysis, deep learning algorithm development, deep learning validation and prediction model development, cloud system development and project management.