Can Medtech Manufacturing Benefit from Deep Learning?

Machines could learn like humans, offering quality and productivity benefits, among others, according to an automation and machine vision expert speaking at MD&M Minneapolis.

Image of a medical screw courtesy of Cognex

Image of a medical screw courtesy of Cognex.

Given medtech’s needs for high levels of quality and precision, machine learning could be put to good use in medical device manufacturing. One particular type of machine learning called “deep learning” could offer particular benefit as it could enable machine vision systems to learn how to recognize and respond to defects much like a human could, according to Bhaskar Ramakrishnan, technical sales engineer, DWFritz Automation.

Ramakrishnan will be speaking about potential manufacturing roles for deep learning in the MD&M Minneapolis October 31 presentation, “Deep Learning Machine Vision for Critical Quality Control.” He will be joined by co-speaker John Petry, marketing director, vision software, Cognex.

Topics to be covered include:

  • Introduction to deep learning.
  • Applications of deep learning in the medical field.
  • Best practices for factory deployment.
  • A case study.

MD+DI asked Ramakrishnan and Petry a few questions about deep learning and their upcoming presentation.

MD+DI: Can you define deep learning? Does it involve machine learning or artificial intelligence?

Ramakrishnan: Artificial intelligence relies on a variety of methods, including machine learning and deep learning, to learn from patterns found in data. Deep learning is a subfield of machine learning that mimics the neural networks in the human brain by creating an artificial neural network (ANN). Like the human brain solving a problem, the software takes inputs, processes them, and generates an output. This method uses weights that are adjusted through a training program to teach the ANN how to properly respond to inputs. So more repetitive teaching makes the ANN stronger and therefore better at identification or prediction. It is like a child learning to recognize the alphabet or multiplication table. Machine learning uses several techniques such as deep learning, regression analysis, Bayesian networks, logic programming, and clustering to implement artificial intelligence into a system.

MD+DI: How has deep learning been applied to date in machine vision?

Petry: The human brain and eyes work well in identifying and classifying defects. However, factors such as fatigue, distractions, and operator-to-operator subjectivity make it difficult to achieve consistent results across production lines or from shift-to-shift.

Machine vision using rule-based algorithms are excellent for part location and gauging applications, but they depend on consistent part appearances. Cosmetic and structural defects by their very nature are unpredictable. Traditional machine vision approaches struggle to distinguish acceptable variations in part appearance from scratches, stains, and other true defects. And even when initially successful, they can be challenging to maintain when confronted with new materials, part changes, or machine-to-machine lighting variations. Deep learning makes this simpler by example-based training. When presented with new conditions, the user adjusts the neural network by adding more examples that illustrate the new conditions. This produces solutions that are much more robust in the face of real world variations and that are easier to maintain over production lifetimes.

MD+DI: How has deep learning been used for medical device manufacturing?

Ramakrishnan: Deep learning has several uses in medical device manufacturing. For example, it is used for package seal inspection of Class 3 devices and finding defects such as scratches on femoral knee implants. It can also identify missing components from a medical set. In addition to defect detection, deep learning can often classify the type of defect, enabling closed-loop process control. And deep learning vision is also used for assembly verification to ensure all components are present in packages such as surgical kits before they’re shipped.

MD+DI: What industry challenges, trends, or unmet needs will deep learning address?

Ramakrishnan: Deep learning can improve quality control in the medical device industry by consistent results across lines, shifts, and factories. It can reduce labor costs, through high-speed automated inspection. And it can help manufacturers avoid costly recalls and resolve product issues, ultimately protecting the health and safety of consumers.

MD+DI: What do you hope that attendees will learn and do differently after hearing your presentation?

Ramakrishnan: Attendees will learn about deep learning applications, how deep learning can be used, and how to apply deep learning effectively for defect inspection. They will discover a cutting-edge way to reduce manufacturing costs while maintaining the highest quality possible.

 

Don't miss the MD&M Minneapolis October 31 presentation, “Deep Learning Machine Vision for Critical Quality Control.” 

Daphne Allen

Daphne Allen is editor-in-chief of MD+DI. She previously served as executive editor of Pharmaceutical & Medical Packaging News, which serves as the pharmaceutical and medical device channel of Packaging Digest. Daphne has covered medical device packaging, labeling, manufacturing, and regulatory issues as well as pharmaceutical packaging for more than 20 years. She is also a member of the Institute of Packaging Professionals's Medical Device Packaging Technical Committee. Follow her on Twitter at @daphneallen.

 

500 characters remaining