Physician Assistance & Training

Screen Shot 2018-01-02 at 12.41.48 PM.png

With the increase in AI and machine learning capabilities, we are making groundbreaking advances. Recently a team at Stanford built a model that outperforms the average radiologist at diagnosing pneumonia from chest X-rays [1]. Another team recently built a deep learning based model that detects skin cancer from photos with an efficacy rate on par with tested Dermatologist experts [2]. And we here at Xyonix are building a number of AI models that help physicians. In this video below, you can see a model of ours in action detecting papillary carcinoma lesions from a cytoscopy (procedure that allows your doctor to examine the lining of your bladder and urethra, or tube that exits urine from your body). Note at 0:21 when the lesion detector activates on seeing the carcinomas (the sponge-like entities). You can also get a glimpse into what parts of the image the AI thinks are important in determining whether a given frame has a lesion present or not.

We at Xyonix are building other models for another company to assist general physicians and cardiologists in diagnosing anomalous heart conditions based on a smartphone based audio signal. For another fortune 500 company, our Xyonix built models are in production today constantly analyzing thousands of recent surgery videos and helping to make surgeons perform better. If you have data that you suspect could help a physician make a more accurate diagnosis or improve, we can rapidly assess your hypothesis and help you get your tool into the hands of physicians where it can make a real impact.

Interested in using AI to analyze your diagnostic or other healthcare data? >> CONTACT US << and let's dig in together. Want to find out more first, then

read our detailed Whitepaper on how AI is changing healthcare -- you will be affected.

REFERENCES

  1. CheXNet: Radiologist-Level Pneumonia Detection on Chest X-Rays with Deep Learning, Rajpurkar et al, rXiv:1711.05225

  2. Dermatologist-level classification of skin cancer with deep neural networks, Esteva et al, Nature 542, 115–118 (02 February 2017) doi:10.1038/nature21056

  3. Photo by Adrian Clark / CC BY