About AI

Artificial Intelligence or AI for an acronym, technically means making a computer eligible to make informed decisions about a problem which are sometimes beyond processing capabilities of a normal person or in most cases too time taking to make relevant impact. Artificial Intelligence helps improve decision making by enabling computers to make real time, high precision impactful decisions. Real life scenarios utilizing Artificial Intelligence are for example Cab Aggregators suggesting high density pick up areas considering area specific details like weather, day, time, local events, etc. to cab drivers to accordingly station their cabs. In fact, we are increasingly interacting with our computers by just talking to them, whether it’s Amazon’s Alexa, Apple’s Siri, Microsoft’s Cortana, or the many voice-responsive features of Google. Chinese search giant Baidu says customers have tripled their use of its speech interfaces in the past 18 months. Then there are the advances in image recognition. Google, Microsoft, Amazon, Baidu all have features that let you search or automatically organize collections of photos with no identifying tags. All these use Artificial Intelligence technologies under the hood.

At PatientMD we are using TensorFlow, an open source Deep Learning framework provided by Google which powers many of Google’s in-house products which includes it most widely used product search result website ranking and its award-winning image recognition model which it uses for many of its internal projects. We use TensorFlow framework, utilizing the Inception-V4 architecture to recognize and classify our skin cancer lesion images.

Our algorithm gets trained on skin cancer lesion images which are a well balanced set of benign and malignant images. The images are well curated, labelled and contains additional facets such as the gender, race, position of the lesion, age of the lesion among other things. The images used have been verified by certified radiologists.

At PatientMD we aim to provide the highest possible accurate prediction for the images uploaded, specifically either benign or malignant. However, we are still in the beta phase wherein our algorithms are continuously getting tuned continuously. We have achieved an accuracy of 90% so far, hence we are still far off from committing to our intended goal which is our prediction results to be used as substitute for radiologist’s physical examination.