Pathologists and clinicians take a lot of time to thoroughly analyse complex images such as tissue sections or X-rays – and they still make mistakes. Even experienced practitioners. Two start-ups from the Charité and BIH have developed AI-based solutions for better diagnostics. Both are supported by Ascenion and the BIH DHA.
Aignostics: Precision diagnostics for pathology
Cancer, infections, autoimmune and other diseases cause visible changes in patients’ tissues. Analysing these changes has been the cornerstone of clinical diagnostics for decades. The AI-platform from Aignostics, developed by Prof. Frederik Klauschen, Prof. Klaus-Robert Müller and their colleagues, is able to recognize known disease-associated characteristics in standardized form to a high degree of precision, increasing throughput while lowering error rates.
But this is not all: the platform helps to identify new markers that predict whether or not a patient will respond to a particular therapy – for example during a clinical trial. The AI-based analysis not only reveals associations between complex tissue images and clinical outcomes, but also identifies the pixels in these images critical for determining this correlation. Based on this information, pathologists can identify markers of prognostic value for the therapeutic approach under investigation.
dentalXrai: Better decisions in dentistry
X-ray images are used in dental diagnostics to detect caries, infections, implants and root fillings. Differentiating between these is not always easy – even for experienced practitioners. Prof. Falk Schwendicke and his team had developed an AI-based platform that can help. Dentists and companies specializing in dental health can simply upload their X-rays for immediate analysis. The results are precise and available in a form that is compatible with software used in dental surgeries. The time saved can be used more efficiently for patient consultation and treatments.
With the help of a vast dental data set which not only includes X-ray images, but also clinical and socio-demographic data from the Charité and further clinical partners all over the world, the underlying model has - and continues to be - optimized.
(Annual Review 2020)
