Deep Cube’s algorithms are able to help diagnose Alzheimer’s disease with 97.3 percent accuracy outperforming Google, the MIT and the traditional diagnosis by doctors. The Lausanne based startup and the University Hospital of Geneva (HUG) have entered a partnership to develop highly efficient AI-based algorithms for accurately diagnosing early-stage Alzheimer's disease.
Deep Cube developed an innovative artificial intelligence (AI) technology-based on Deep Learning. The technology focuses on two core areas; analyzing skin disorders and diagnosing human brain PET scans for Alzheimer’s disease. According to Deep Cube, their new AI diagnostics could also play a key role in measuring drug efficacy in Alzheimer’s clinical trials.
Together with the University Hospital of Geneva (HUG), Deep Cube will use its AI-based algorithms to better evaluate brain positron emission tomography (PET) scans measuring glucose metabolism to diagnose patients with Alzheimer’s disease.
Alexandre Gouy, Deep Cube’s AI engineer on Alzheimer’s, commented, “Accurate early diagnosis is crucial in neurodegenerative diseases as it allows therapy to begin earlier. Additional benefits of the AI Model are to improve Alzheimer’s diagnostics at hospitals. Our neural networks architectures combined with imaging and biomarkers reach state-of-the-art AI performance using a minimal amount of data, approximately 200 PET scan images.”
Outperforming Google
Deep Cube CEO Chris Patris de Broe also commented, “The future of AI in neurosciences is a combination of three input sources: medical imaging, biomarkers and genomic data. At 97.3 percent accuracy, our AI team has been able to surpass two other Alzheimer’s AI diagnostic models, one at Google with 94.2 percent sensitivity and another at the Massachusetts Institute of Technology (MIT) with 85 percent sensitivity. Even if data is not comparable, in each of those AI models the results are tangible. So far, the best doctors have been able to do in analyzing Alzheimer’s PET images is 93 percent accuracy, that is, 7 percent misdiagnoses. Our Deep Cube AI Alzheimer’s Model is demonstrating minimal misdiagnoses at 2.7 percent, significantly lower than human interpretations.”
(Press release)
Please login or sign up to comment.
Commenting guidelines