The last few years have seen a number of sites emerge that offer a second opinion service for our medical conditions.  The likes of CrowdMed and Figure1 allow us to upload details of our condition and receive a second opinion from either medical professionals or simply other people from around the world.

These services all rely on human to human interactions however.  A Californian company are looking for test images to put their cancer recognition software through its paces.  They have sent out a call for women to donate their mammogram images to help them test software that they believe will provide a breakthrough in early detection of malignant tumors.

Requesting all mammograms

The company’s Lesion Tissue Profiling technology has been tested over the last three years.  The technology aims to detect malignant tumors that are not visible to the human eye.  It works because such tumors typically take on a number of basic shapes that distinguish it from healthy tissue.

A database of mammographic images has been created to allow the company to quickly and accurately compare billions of tissue profiles.  The data within each image is highlighted using a color illustrator, with the shape found then cataloged to provide a tissue profile that can be compared with other profiles.

When these are compared with healthy mammograms, the company reveal that they can accurately spot precursor tissue profiles in both 2D and 3D images.

Calling the crowd

The next stage in the development of the product is to train the technology on as many mammogram images as it can find.  The company has therefore asked women that have been diagnosed with breast cancer to donate their mammograms to allow comparison to be made as the disease progresses within them.

They’re hoping to ‘recruit’ mammograms from the whole duration of the process to try and compare earlier mammograms with those taken later on when diagnoses were confirmed.

The company believe that the larger their repository of mammogram images, the more accurate their profiling technology will be at detecting tumors at an early stage.

The key to all of this however is scale.  Traditionally clinical trials and medical research can call upon data from maybe a few thousand patients.  Larger studies don’t tend to use a vast collection of individual images, which is why crowdsourcing can be so valuable.

It’s certainly an interesting project with a very worthwhile goal.  If you’d like to contribute, you can do so via their website at http://www.YourScan.org