The reality of AI for medical applications
I vividly remember watching Google's DeepMind AI beat ATARI games and discovering the tricks we all tried as kids. After this breakthrough, their team amazingly beat one of the world's leading professional GO players - Lee Sedol. And GO is a game that, because of its huge search space (unlike chess), couldn't be solved by algorithms with clever heuristics. Instead, real learned intuition and human-like intelligence were required.
-> I highly recommend watching the AlphaGo documentary here <-
All that was some years ago. Since then, AI has rapidly advanced from research papers to real-world applications - in the consumer sector, but most productively and beneficially in the medical field. Here, developers are training machines not to play board games, but to augment medical images and help with diagnostics.
AI reimagines medical imaging
The medical industry is experiencing a major technological leap thanks to AI. Far from currently unrealistic Artificial General Intelligence (AGI), this field is an early adopter when it comes to AI augmenting diagnostics in medical imaging such as X-rays, MRIs, and mobile ultrasounds.
AI can reconstruct images to reduce patients’ time in the MRI machine, enhancing resolution before medical professionals view them. Clearer images in less time not only means more patients can be scanned; it also makes it easier for medical professionals to make a diagnosis.
Another use case is the enhancement of the diagnostics process itself. It takes many hours of training before a medical professional can derive an adequate diagnosis from complex ultrasound or MRI images. AI is trained on an amount of data that a human could never look at in a lifetime. Moreover, its training never ends. However, diagnosis is not outsourced completely to the machine - its task is the augmentation and pre-filtering of images, as well as the confirmation of professional diagnostic opinions on image data. Imagine AI saying, "Hey, look at this shadow in this image. Could this be a tumor?" - Yes, that’s the reality we already live in.
x86 cores + graphics cores + AI cores = Intel Core
To make AI inference broadly available, more and more silicon vendors are adapting their portfolios to incorporate enhanced AI functionality in their compute engines. Take the first generation Intel Core Ultra processors for example: They now also feature a Neural Processing Unit (NPU), an ultra-efficient edge AI inference accelerator, significantly more powerful graphics, enhanced GPGPU performance, and 10-year availability. In detail, the new NPU in the Meteor Lake processors executes machine learning algorithms and AI inferences approximately 20 times more power-efficiently than x86 instruction sets. And with NPU performance reaching the level of discrete GPUs, 1.9 times faster graphics and/or GPGPU processing results in a more detailed and immersive user experience compared to the previous Intel Core generation.
Old COM out + new COM in = AI capability
The best part? With standardized Computer-on-Modules (COMs), especially COM Express, you can implement the new AI capabilities easily, without needing to change your existing designs. This is exactly why industries in rapidly evolving markets are designing their products with COMs. As the AI trend and its use cases evolve, the flexibility of a COM and carrier board solution gives designers the freedom to adapt their products to rising computational needs in two simple steps: Unplug the previous module, plug in the new one, and off you go…
This reduces the total cost of ownership of your applications, as in many cases there is no need to integrate external AI acceleration cards.
Standardized Computer-on-Modules like the conga-TC700 COM Express Compact are an ideal platform to upgrade medical applications with the latest AI capabilities.
Find details here: https://www.congatec.com/en/products/com-express-type-6/conga-tc700/
Conclusion
AI excels at spotting patterns and anomalies that even professionals might miss. Medical experts are often under a lot of pressure. Anything that can speed up diagnosis and focus attention on the critical spots is more than helpful. I noticed this when I showed the MRI of my finger that was bothering me to three different doctors and received three different opinions. So, I – for one – am positive about this technology. How about you?
To learn more about the advantages of the Intel Core Ultra technology with integrated NPU,
please visit: https://www.congatec.com/en/technologies/intel-meteor-lake-h-based-computer-on-modules/
