Breast cancer: earlier diagnosis with AI
Breast cancer: Deep learning could enable earlier diagnosis
Detecting breast cancer as early as possible is a goal for many healthcare professionals. The earlier treatment begins, the more lives can be saved. With support from EuroCC Austria, Austrian start-up Salataris AI is developing a deep learning model that supports medical experts in identifying tumours at an early stage. The model is trained on mammography and MRI images to detect even small and rare tumours.
Bettina Benesch
Breast cancer is the most common form of cancer among women – and although men are affected far less frequently (around one percent of cases), they are not exempt. In Europe, approximately 660,000 people are diagnosed with a malignant breast tumour each year – about the same as the population of Luxembourg. Every year, 156,000 people die from the disease – roughly equivalent to the population of Heidelberg, Germany. All this is caused by just a few rogue cells. If they could be detected earlier and more accurately, thousands of lives could be saved.
This is exactly where Salataris AI aims to contribute: supported by EuroCC Austria, the team is developing an AI model to detect breast cancer in mammography and MRI (magnetic resonance imaging) scans. The first step is to improve the image data. Mammography and MRI images often contain so-called artefacts – distortions in the image, such as speckled patterns or fine-grained noise, which make diagnosis more difficult. In order to reliably identify altered cells and tumours, this noise must be removed and both image quality and contrast enhanced.
Fewer false positives and negatives thanks to AI
The improved image quality enables the AI model to deliver better results and thus helps reduce the risk of false negative and false positive outcomes. A false negative occurs when a tumour is present but goes undetected – meaning it is overlooked. A false positive, by contrast, is when a tumour is diagnosed even though none exists. In both cases, patients suffer: a false negative test result gives a false sense of security, and a false positive typically leads to a biopsy, which takes time, costs money and causes anxiety – all of which could have been avoided in the case of a truly negative result.
Integrating AI into hospital IT systems for better diagnostics
Salataris AI uses deep learning, a subfield of artificial intelligence, and is currently in the training phase of its model. It is being trained on large volumes of mammography and MRI scans to learn how to detect tumours reliably. By the end of the project, the model should be easy to integrate into other software systems via an API (application programming interface). It is also designed to run directly on compact, powerful hardware – such as an NVIDIA Jetson (a small computer optimised for AI) or a Field Programmable Gate Array (FPGA), a highly flexible, programmable chip. In tThis way, it can be easily embedded into hospital IT systems.
At the end of the project, the diagnostic tool is intended to be easily integrated into other software via a digital interface. This would allow hospital software to be extended with a new function.
Different learning approaches improve model performance
During training, the aim is to test different models and continuously optimise their learning performance. This process is known as hyperparameter optimisation (HPO): developers search for the best possible configuration of the model and its training method. Since this involves a great deal of trial and error, significant computing power is required – which is exactly what a supercomputer provides. It allows many configurations to be tested in parallel, with the best-performing model selected at the end.
The team uses several learning approaches at once, including:
Unsupervised learning algorithms to remove image noise, enhance image quality and detect hidden features in rare or abnormal cases. Unsupervised learning means the model identifies patterns in the data without being told the correct answers – for example, whether dense tissue indicates a tumour.
Supervised and semi-supervised learning: Supervised learning uses labelled data (e.g. “this image contains a tumour”) to train the model. Semi-supervised learning combines both labelled and unlabelled data.
Manifold learning, which searches for hidden structures in complex datasets. A manifold is a mathematical structure that helps explain how complex, high-dimensional data may actually lie on a smooth surface – like a drawing on a crumpled sheet of paper: a two-dimensional surface distorted in space. In tumour detection, this means that MRI scans with similar features – like tumour boundaries – may lie close together on the same part of the manifold. The model tries to learn this structure to better identify changes in tissue.
Contrastive learning, which helps the model distinguish clearly between similar and dissimilar cases.
Self-adaptive learning, which allows the model to continuously adapt to new data and changing conditions. It keeps learning during operation and can use any learning type – supervised, semi-supervised, or unsupervised. This helps improve the model’s accuracy over time.
Reinforcement learning via feedback: the AI system receives positive reinforcement for correct decisions and negative feedback for incorrect ones. This also improves its ability to recognise tumours – and to distinguish artefacts from true findings.
Artificial intelligence for image recognition always requires substantial computing power. A supercomputer is ideal for this: it consists of many interconnected machines and can easily handle the training of several hundred thousand images.
AI and supercomputing: faster and better results
Salataris AI is training its model on the LEONARDO supercomputer located in Italy, accessed via EuroCC Austria. Training an AI system for image recognition always requires large-scale computing power, as the image files are often large and the model needs thousands of images to learn patterns. This is where high-performance computing (HPC) comes into play: HPC systems consist of many interconnected computers working together, making them significantly more powerful than any individual machine. A supercomputer can easily handle the training of a model with hundreds of thousands of images. Salataris AI also wanted to understand HPC workflows and learn how to scale AI training on such systems.
To work efficiently across multiple interconnected computers, a process called parallelisation is used – dividing a large computing task into many smaller ones that are processed simultaneously by high-performance GPUs. This only works smoothly if the right tools and techniques are applied. That’s why EuroCC Austria offers expert support to start-ups like Salataris AI entering the world of HPC.
Image enhancement with deep learning: find out more in part two
The second part of this success story will explore how mammography and MRI images can be enhanced – with and without AI – and how Salataris AI is using deep learning to do exactly that. Available here from 9 June 2025.