Computer Vision News - January 2022

54 MedTech Application with a clinical supervisor, MihneaGangal , a urogynecologist and surgeon at Notre Dame Hospital, from the outset. “ We sat down with him and talked through the unmet need and how we could solve it, ” Negin tells us. “ He has lots of surgeries for pelvic organ prolapse as well as pessary fitting. Based on his experience, the process of being fitted with a good pessary is trial and error. For some patients, it’s not a good fit and they come back wanting to have surgery because it’s so uncomfortable, or they have a problem with displacement, so they get fed up and discontinue using it. ” Currently, for measurement of the vaginal canal, doctors use their hands and an interactive assessment tool called POP-Q . The team had been developing a device to do this – they even built a prototype and tested it on a cadaver – but through research and development, found theycoulduseequations to translate the data from doctors to fit their need. When a patient goes to the doctor, the doctor validates their measurements and inputs them into FemTherapeutics’s software. With those measurements, and the use of AI, the team can design an optimal pessary for the patient and send it off for production. They are using 3D printing as it is simple, fast, accurate, and accessible and they can ship the pessary directly to the patient afterwards. The model involves machine learning and AI methods . The first part uses imaging to predetermine the design and shape, with two datasets of CT scan images and MRIs . “ We annotate the images one by one to detect the parts we want, ” Negin explains. “ We use Gaussian curvature, deep learning-based segmentation methods, and contour and corner detection . We make sure we do it for all slices of the CT scans and MRIs and use software like Fiji to put all the MRIs together to make a 3D shape of the segments we need. We use segmentation and localization to make sure we accurately detect the required parts. We bring this data into design and simulation