Shock due to trauma, infection and heart attack affects 10 million patients in USA and EU annually. Prof. Johansson at Rigshospitalet has identified a subgroup (20%) of the shocked patients having TOXic catecholamine release to the blood resulting in acute heart failure with >80% mortality rate. The shocked TOX patients can be identified by the biomarker succinic acid > 10nM in the blood enabling targeted treatment with beta1-selective blocker, which according to published observational data, will reduce mortality by 50%. A PCT application has been filed by Rigshospitalet concerning this[WO2022/233994 A1]. The TOX project will validate the 10nM cut-off level of succinic acid to identify TOX in 500 patients with septic shock by mass spectrometry using an existing biobank at Rigshospitalet. A point-of-care biosensor will be developed at DTU Bioengineering that measure succinic acid in plasma in the required nM range to be used in a proof-of-concept trial in 200 TOX patients with septic shock evaluating the effect of beta1-selective blockers vs. placebo on 30-day mortality conducted at Copenhagen University Hospitals. The result of the trial including a Health Technology Assessment conducted by VIVE will be used to implement the TOX test globally through license agreements, partnering or take-over with global diagnostic companies, and MoxieLab ApS, a spin-out from Rigshospitalet, now owning the IP. Roche Diagnostics has already shown interest in the technology.

Speech technology (ST) is a growing industry, with an estimated annual compound growth of 16-19% in the next five years. The majority of ST development is targeted towards high-resource languages, such as English. Danish, as a low-resource language, risks falling behind technologically and missing revenue from this technology. Modern ST is based on ML algorithms, which require large amounts of data. Lack of Danish data is causing Danish ST to fall behind. The CoRal project will remedy this, by creating the first large-scale high-quality dataset consisting of 1,500 hours of transcribed Danish speech, both conversational and read-aloud, comparable in size to English state-of-the-art speech datasets, with eight models, incl. speech recognition and text-to-speech models. The dataset and models will be publicly available and enable use cases which were previously impossible even for some of the largest Danish companies e.g., ATP. Publicly available materials will be developed which will help companies implement the technology and allow SMEs to enter the area with minimal risk, as they can use models and data with almost no prior expertise. In addition, the project will ensure that the dataset represents all variants of spoken Danish, across regiolects, ethnolects, etc. This will allow authorities to provide speech services to all citizens, even marginalized demographics. The scale and diversity of the dataset will push Danish ST up to the highest standard commercially.