Nanoparticle Tracking Analysis with Confidence! – The use of Machine Learning with the NanoSight Pro
Why Trust Humans?
Humans by design are fallible. Even those of us considered to have great minds can slip up and say, lose a space probe costing more than $450 million dollars due to inconsistencies in calculation measurement units (Mars climate orbiter incident).
When it comes to drug development, the stakes can be just as high. Human error can cost millions in development or more importantly, adverse health effects. So why leave data processing to error-prone humans? This is where machine learning shines! Through statistical methods and repeated exposure to related data, algorithms are trained to identify patterns and make predictions in a fraction of the time without large error margins.
Making the Invisible Visible
We don’t realize how much humans and their wellbeing are dependent on small particles that are invisible to our eyes (Figure 1). The bioparticles that regulate the functions of our bodies can, at the same time, cause human diseases. Pharmaceuticals and several industries involving advanced materials are also impacted by the nanoworld. Drug interactions happen on the cellular scale but can also have profound (and sometimes adverse) effects. Other industries involving nanobubble generation for various applications require established and controlled protocols.
Figure 1: NanoSight NTA particle visualization. NanoSight NTA collects clips of the light scatter from nanoscale particles moving under Brownian motion (white dots).
How Can You Better Understand These Tiny Particles?
Figure 2: Example of NS Xplorer software tracking nanoparticles. Nanoparticle movement is tracked particle – by – particle.
Malvern Panalytical’s NanoSight can detect very small biologics, bioparticles, nanobubbles and various other synthetic nanoparticles using Nanoparticle Tracking Analysis (NTA). NTA utilizes light scatter to illuminate and track the Brownian motion of nanoscale particles (Figure 2). In return, users can very quickly understand the complex nature of particles including their size and concentration.
You can imagine that ‘catching’ all the particles in the first place is critical for accurate and precise particle tracking. NTA users might find themselves often asking, “Is this a particle or not?”, “which setting is the best?” and then have to adjust the Detection Threshold parameter, or simply forget to do so. What if these parameters could be precise and automated? Enter Machine Learning with the new NS Xplorer software.
Machine Learning Empowers the New NS Xplorer Software
Malvern Panalytical introduced the NanoSight Pro earlier this summer in 2023. Nanoparticle Tracking Analysis has never been so quick, easy, and accurate!
Machine learning can now accurately mark the center of the dancing dots as part of the NTA tracking algorithms. This includes capturing the very dim particles affording more accurate analysis of your sample while you are simultaneously running further workflows! This further removes confirmation bias and mitigates human error.
The Particle detection Neural Network Model (NN) used with the NanoSight Pro was trained over weeks on thousands of NTA images with various levels of complexity. Data was manually assessed and confirmed by human counterparts for its precision and accuracy through the development process (Figure 3).
Figure 3: NanoSight particle detection Neural Network structure. The data output by the neural network was validated to determine if the model correctly identified particles.
Validation of the model was carried out by benchmarking against typical and expert NTA users. The data in Figure 4 clearly shows that the machine learning (Neural Network model, NN) precisely identified particles for further use by the tracking algorithms. The NN removes subjectivity and delivers more repeatable and accurate data than even the most experienced users!
Figure 4: Comparing particle identification by machine learning and NTA standard and expert users. The Neural Network models ability to detect particles outperformed not only typical users, but even the most experienced users (NN vs Perfect User vs Typical User).
Now thanks to machine learning, not only do you save time setting processing parameters, but you don’t even notice when the data is being processed…Eureka! NTA processing is now automated, speeding up analysis and giving the user more freedom to analyze additional samples, setup further protocols, or simply have a nice coffee break!
Applications
The new NanoSight Pro can help you measure a wide range of particle types including the following:
Now, relax and let NanoSight NS Xplorer characterize your particle systems and accurately report their size and concentration!
Want to learn more? Join our upcoming webinar:
