HMN 2026: How Engineers test, validate novel method to improve pharmaceutical research and development

A patent-pending innovation created and validated in Purdue University’s College of Engineering could strengthen pharmaceutical research and development in the areas of batch verification, encapsulation efficiency screening and regulatory compliance workflows.

Mechanical engineering professor Arezoo Ardekani and her doctoral student Kaeul Lim have leveraged hyperspectral imaging (HSI) and machine learning (ML) to develop a novel, label-free, noninvasive method to characterize nanoparticles.

The method enables robust, scalable classification of liposomal drug carriers and lipid nanoparticles.

“It provides real-time, nondestructive quality control during nanoparticle formulation and manufacturing,” Ardekani said. “Due to its noninvasive nature, it can be adapted into industrial high-throughput analytical platforms.”

Ardekani said the Purdue method’s classification accuracy for nanoparticle type approached 99% under optimal parameter conditions.

Ardekani’s research was published in the International Journal of Pharmaceutics. She disclosed the innovation to the Purdue Innovates Office of Technology Commercialization, which applied for a patent with the U.S. Patent and Trademark Office to protect the intellectual property.

Challenges of traditional analytical methods

Ardekani said label-free characterization of nanoscale drug delivery systems remains a critical challenge in pharmaceutical research.

“Traditional analytical methods are labor-intensive, low-throughput or require labeling, which can interfere with nanoparticle functionality,” she said.

The Purdue method focuses on enhancing image quality within HSI data. Ardekani said its ML capabilities significantly reduce classification complexity and improve overall accuracy.

“Traditionally, HSI’s application to nanoparticle analysis is limited,” she said. “Data is often lost to noise and overlapping information.”

How the Purdue method works

Liposomal formulations are deposited onto precleaned glass microscope slides and imaged without further modification or staining.

“The HSI system captures scattered light spectra from a sample through line-by-line spatial scanning, where each pixel’s information represents the spectrum at that location,” Ardekani said. “Hyperspectral images are recorded by using an enhanced dark-field illumination system attached to a microscope.”

Ardekani said following the image preprocessing phase, classification is performed using convolutional neural networks.

“The convolutional layers learn important features from the hyperspectral data,” she said. “Those features are then passed through the fully connected layer and the output layer to classify the nanoparticles.”

The content is provided for information purposes only.