Computer Vision News - September 2020

Paige 15 whom Paige can bring added value by helping them to digitize more labs and hospitals. Looking to the future, Carla thinks the landscape will look very different in five years’ time because Paige is not the only company in this space and the compelling reasons to go digital will continue to grow. “Right now, particularly in the US, there are very few teams of pathologists that are diagnosing digitally. Most teams are still using microscopes. Once there are multiple reliable, FDA-cleared, broadly available products, trusted by key opinion leaders, many more practices will be persuaded to go digital. It is really going to transform how pathology looks day to day. If a company creates a product that is so irresistible that it convinces all stakeholders to do it, or a digital diagnostic that does something that nothing else can do, it will have a huge advantage.” Reimbursement is also an important factor in terms of getting a product to market as it can take a long time for technologies to be reimbursed by insurance. Carla tells us small companies must think carefully about this because without reimbursement they need to find private insurers and hospitals that are willing to pay for these products without Medicare/Medicaid paying for them. You can only give your product away for free so many times before you go bankrupt. Therefore, medical device start-ups tend to raise bigger rounds than tech start-ups. You need a much longer runway so that you can get to that place where you have regulatory clearance and reimbursement, or a business model that does not need reimbursement and keeps the company sustainable. Finally, Carla says Paige is lucky to have so many presentations of rare cancers due to its ties to MSK, but a repository of rare presentations of disease would be so helpful as without it, there are always things that the algorithm will miss. “We train our systems on tens of thousands of slides, but most other companies in this space are an order of magnitude lower and we are pretty convinced that they are not going to see the entire range of normal and the entire range of cancer,” she laments. What is most needed is multiparametric data : the CT, the biopsy, the genomics, the driver, the outcome. Those five pieces of data are in five different repositories in the hospital and curating a dataset with all five for enough patients is extremely difficult and expensive, but you could do incredibly interesting things with that.” If all this has piqued your interest, you will be pleased to know that Paige is hiring!