AI in medicine helps specialists focus on other tasks

When ultrasound was developed over 75 years ago, it was based on a natural mode, specifically bats that scan their surroundings using sound and build up an image using their sense of hearing.¹. Modern medicine is again modeling itself on nature. However, the focus is not so much on sensory perception as it is on the body's  processing organ: the brain or the use of neural mechanisms in artificial intelligence (AI). “AI broadens our competence as physicians and provides a basis for making decisions. The aim here is not for AI to operate independently,” says Prof. Dr. Mathias Goyen, Chief Medical Officer Europe at GE Healthcare, who is actively involved in the opportunities emerging in medicine through AI.

The interest in artificial intelligence is growing

“When it comes to examining data quickly and according to analyze precisely defined criteria, AI is far superior to humans in terms of speed,” says Goyen. AI is primarily used in medicine to analyze two data types: diagnostic image acquisition and genetic data². Computer specialists often use machine learning for this. The computer learns to identify illnesses based on diagnostic data – as a physician would learn in their training. This enables the computer to provide the physician with a basis for making a decision. On average, a hospital produces 50 petabytes, i.e. 50,000 terabytes of data per year. However, only ³ percent of it is used3. Goyen is convinced: “Artificial intelligence can change this.”

Current publications indicate that the interest in AI is continuously growing in many areas of medicine⁴. This includes emergency medicine, where AI-assisted systems can help make vital diagnoses almost in real-time. It is also gaining importance in general medicine, cardiology, and gynecology. This is due to an important feature of AI: learning and interpreting new situations. “That's why AI is so valuable in medicine. It helps us doctors make better use of our limited time.” The computer relieves the medical specialists by taking up repetitive tasks.

Goyen adds: “With the AI in our devices, we want to free up the capacity of physicians for important tasks like direct patient contact.” Although this will change current job profiles, the human aspect of care will be extremely important and will remain in human hands. The computer works in the background to support an ever-increasing number of tasks. “Artificial intelligence in medicine is not going to go away, but will rather continue to bring us forward.”

AI in medicine: evaluation of ultrasound recordings

For over 20 years, the area of ultrasound diagnostics, in particular, has witnessed efforts to use artificial intelligence, such as for detecting liver diseases like hepatitis or cirrhosis⁵. Another use is for classifying nodules in the breast or tumors in prostate tissue^6,7. Neural networks in computers were used for this early on – they are based on a combination of processing units that work similar to neurons and detect patterns in large data records. With more powerful processors, computer scientists could develop this method further. This helps when little data is available for “training” a neural network⁸. Sophisticated networks help computers solve increasingly complex problems independently. However, for it to be approved⁹, a “diagnosis program” must deliver reproducible results with consistently high quality. To ensure this, an algorithm can be generated as soon as the training phase is completed, i.e. a formalized schema that is then followed to analyze data.

To broaden the use of AI in medicine, connectivity and flexibility of the devices and computer systems are essential. This is where the Edison platform of GE Healthcare comes in. This system consolidates hospital data from multiple sources that can be analyzed using various apps. The latest generation of smart diagnostic devices from GE Healthcare is also integrated with the system. The various modules – some of which originate from other technology companies – allow the platform to be individually adapted to any hospital. Because one thing is certain, says Goyen: “Artificial intelligence will play a big role in the future, but we need to decide today: do we want to help shape this future or follow in the footsteps of pioneers from industry and medicine who have already broken new ground.”

¹ Kang et al., 2012: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3512173/
² Galimova et al., 2019: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6449768/
³ GE Reports: https://www.ge.com/reports/ai-healthcare-expert-doctors-machines-make-brilliant-match/
⁴ Galimova et al., 2019: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6449768/
⁵ Ogawa et al., 1998: https://ieeexplore.ieee.org/document/737666
⁶ Joo et al., 2004: https://ieeexplore.ieee.org/document/1403434
⁷ Loch et al., 1999: https://onlinelibrary.wiley.com/doi/abs/10.1002/%28SICI%291097-0045%2819990515%2939%3A3%3C198%3A%3AAID-PROS8%3E3.0.CO%3B2-X
⁸ Ravishankar et al., 2017: https://arxiv.org/pdf/1704.06040v1.pdf
⁹ Ernst & Young, 2018: https://www.ey.com/Publication/vwLUAssets/ey-digitalisierung-algorithmisierung-und-kuenstliche-intelligenz-im-pharmarecht/$FILE/ey-digitalisierung-algorithmisierung-und-kuenstliche-intelligenz-im-pharmarecht.pdf