Computer Vision News - January 2018

This feature could also alert a supervisor if the nurse needs help with certain manual aspects of the procedure. In addition to this, they want to take annotations and refine them. A huge variability exists in the proficiency of the nurses screening for cervical cancer. Some nurses get the answer right most of the time while others only get it a small part of the time. This makes it difficult to know which labels to trust. Labeling the data to train a set later makes it even more complicated. They need to actually take the images and send them through a sort of QA process , not only for image quality, but also for the decision that was made. They need to evaluate both aspects of the data before using the annotation and the label for machine learning . The third issue is that the clinicians do not always trust another clinician’s opinion, but they also refer to the golden standard of taking a biopsy. It is difficult to perform a biopsy in low resource settings for the same reason that it’s difficult to do a pap in a low resource setting. It requires some infrastructure, a lab, and someone to process the tissue. The tissue goes then to a lab where it is dehydrated and put it into a paraffin. Then, they slice small sections of it and add ink before sending it to a pathologist. This pathologist does another subjective step that also needs automation. That whole process does not exist in some countries in sub-Saharan Africa . MobileODT has an office in Nairobi where David learned that although East Africa has several hospitals, they really only reach a small fraction of the patients. With a pathologist, that problem gets even worse. A country like Botswana has only four pathologists to handle every disease. This creates an incredible shortage in solving these kinds of problems. David believes that AI has amazing potential to take on some of these challenges. It could give nurses at the point of care an answer or indication, and even tell them exactly where to look. Even something like that would make a huge impact in healthcare. David insists: “ We want to save as many lives as quickly as possible. That’s why we picked cervical cancer and other diseases because that’s one way we can do just that! “ No matter where they live, women around the world care about their health and want the ability to monitor themselves and stay healthy in order to take care of their children and family. “ We’re here to actually help bridge the gap - David concludes - because there are 5 billion people with access to smartphones, but without access to a physician. That’s just unfortunate, and we need to address it. ” Computer Vision News 13 MobileODT “A huge variability exists in the proficiency of the nurses screening for cervical cancer” Application