Artificial intelligence enables faster and more accurate detection of cancer, which supports the work of doctors and makes it easier to provide personalised treatment for patients.
By analysing images of a biopsy, Ibex's Galen system is able to provide the Gleason score, the size of the tumour and other morphological characteristics. This AI is based on data collected from more than 10 million images of biopsy slides and on clinical research data from institutes around the world. The database also includes statistically very rare cases. In addition, pathologists have annotated more than 150,000 images, to further enhance the quality of the data used. As an artificial intelligence, the model is continually improving, becoming more efficient and more accurate.
The accuracy of the Ibex algorithm has been clinically validated. An external study confirmed its results with an AUC of 0.991. The algorithm also distinguished between low-grade cancers (Gleason score 6) and high-grade cancers (Gleason score 7-10), with an AUC of 0.941. Performance was also excellent in detecting Gleason patterns and perineural invasion, with AUCs of 0.971 and 0.957 respectively.
In clinical practice, the algorithm has been integrated by several diagnostic laboratories and has already analysed more than 11,000 biopsy slides. The alerts generated have enabled clinically relevant cases to be detected, helping to improve the accuracy and speed of diagnosis. The time required to make a diagnosis using AI is 27% less than that required by microscopic analysis. The rate of discordance is also significantly lower with AI (4.84%) than with microscopy (7.13%).
Translated from the original French version by our esanum.fr partners.