AI in Healthcare

Hey everyone,

I'm happy to announce our October event, just after the holidays end on Oct 19th. In this event, we'll host Healthy.io, and Theator research teams to each describe examples of their recent work.

Healthy.io's team will present work on accurate measurements of chronic wounds are required for determining the optimal treatment. and Theator's team will present a multi-instance learning approach for critical view of safety detection in laparoscopic cholecystectomy.

About Yariv[Theator]:
Yariv Colbeci is a Computer Vision Engineer in Theator. As part of his work, Yariv research and develops deep learning based algorithms that enable Theator's Surgical Intelligence platform. Prior to joining Theator, Yariv was part of another Israeli start-up and serverd in 81 unit.He is passionate about AI, Machine Learning and sofware development .Yariv holds a B.Sc. in Computer Science from Hebrew University of Jerusalem.
As part of his hobbies, Yariv likes sailing and skiing

About Shir[Healthy.io]:
Shir Barzel is a Computer Vision and Machine Learning Algorithm Developer in Healthy.io. As part of his work, Shir develops deep learning based algorithms with the aim of producing clinical results that affect patients worldwide.
Prior to joining Healthy.io, Shir was part of several Israeli start-ups with a wide experience of projects ranging from the first personal-autonomous robot to augmented reality operated drones.
He is passionate about using cutting-edge technologies for improving people's health and enthusiastic about the use of brain-inspired deep learning algorithms.
Shir holds a B.Sc. in Electrical and Computer Engineering from BGU and is currently working on his M.Sc. thesis in TAU.As part of his hobbies, Shir likes to play beach volleyball and to swim.

About Sivan [Healthy.io]:
Sivan Biham is a Computer Vision Researcher and Algorithm Developer in Healthy.io, where she works on healthcare-related products.
Sivan holds an M.Sc. in computer science from Weizmann Institute with a specialization in Computer Vision and Deep Learning, and a B.Sc. in both Computer Science and Neuroscience from Bar Ilan University.She is enthusiastic about using her algorithmic skills and knowledge for improving people's health and life. As part of her daily work, Sivan places emphasis on designing and implementing maintainable and modular algorithms, with software architecture principles in mind.
In her spare time, she loves to run and practice yoga.
---
Accurate measurements of chronic wounds are required for determining the optimal treatment.To achieve such measurements we use algorithms for 3d reconstruction. The common approaches for solving the 3d reconstruction problem require labeled data, but labeled data is hard to achieve, even more in medical scenarios. So what can we do if we don't have this labeled data?In this talk, we will present two frameworks that provide a way to handle the lack of labeled data. The first framework is a way to create ground truth for supervised training, even though it is hard to generate it. The second framework is based on a self-supervised method where it enables adding additional information which compensates for the lack of ground truth. Using the above, we will show two examples for training deep learning models which create a 3d model from a single image.
---
Surgical procedures have a clear designated goal, which makes the art of performing surgery a task-oriented action. The performing surgeon follows specific workflow steps that describe the actions needed to reach the surgery goal. In ectomy procedures, such as Cholecystectomy and Appendectomy, the goal is to dissect and remove a specific organ. Safety measures are set to prevent injuries, and the surgeon needs to follow protective methods to avoid misidentification. In Laparoscopic Cholecystectomy (LC), this method is known as Critical View of Safety (CVS). This work illustrates that machine learning can detect CVS accurately enough to be used routinely in the clinical setting, both for educational purposes and in other assessment scenarios. We formulate CVS detection as a supervised Multi Instance Learning (MIL) problem and propose an attention-based MIL model that is trained and evaluated on more than 2,000 surgical videos. It achieves 82.6% mean unweighted accuracy in detecting LC CVS criteria and 84.2% accuracy in the final task of CVS detection.