Medical Image Processing and Meandering as an Entrepreneur

I came to Georgia Tech originally to pursue a graduate degree, mainly because I had no idea how to start a company as I was exiting my undergraduate program. I had tried several ideas as an undergrad that all failed.

While at Georgia Tech, I focused my education on medical image processing and computer vision. Ultimately, I decided that my particular area of research was not particularly commercially viable. Luckily, tangentially related areas became valuable and that is ArrayFire today. In the next few posts, I will describe medical image processing, because that is how I got my start and because it is generally interesting.

MRI, CT, PET, and other medical imaging systems produce anatomical pictures that look inside the human body. Radiologists are the specialists that are trained to interpret these images and give diagnoses and prescriptions based on the imaging information, e.g. the presence or absence of tumors.

However, the medical imaging systems of the world produce more images each day than there are radiologists available to look at them effectively. Medical imaging software accomplishes two primary goals:

1. It aids radiologists in becoming more efficient and effective in their roles
2. In some cases, the software is able to discover information about the imagery that a radiologist is unable to see by eyesight, e.g. the frequency of a signal or texture patterns

Medical imaging data is typically 3D, because the human body is 3D. The medical imaging data can become 4D if a time lapse video is taken of the 3D anatomy.

The large amount of 3D or 4D data produced by these medical imaging systems means that the software must be efficient and run quickly in order to process the large data in tractable amounts of time.

ArrayFire was originally started to build faster code to solve this problem in medical imaging. With ArrayFire, medical imaging software is able to run much faster than it would run without the acceleration, often 10X faster or more.

Luckily, it turns out that the things we had to do to accelerate medical imaging software are equally applicable to all sorts of other engineering, science, and financial computing domains. So our company is much broader than the medical imaging market.

I am always interested to hear how entrepreneurs meander from what they did pre-entrepreneurship, to how they started their company, to how they ultimately find entrepreneurial success. Oftentimes, where they end up is very different than where they began.

How have your meandered through your entrepreneurial career?