Development Of An Automatic Image Analysis Approach To Determine Particle Distribution For Deep Learning Dataset

Student Classification


Faculty Mentor

Dr. Mohammad A. Azad, Chemical, Biological and Bioengineering Department, North Carolina Agricultural and Technical State University, NC; Dr. Tariq Arif, Mechanical Engineering Department, Weber State University, UT


Chemical Engineering

Document Type


Publication Date

Spring 4-2021


This study aims to develop a computer algorithm to determine the particle distribution/homogeneity in a powder mixture from microscopic images. This is an intermediate step for a quick and automatic image segmentation process (typically manual and laborious) using deep learning and will contribute to fast pharmaceutical drug product development through image-based mixture analysis. In this study, microscopic images of binary particles (Ibuprofen or Ciprofloxacin as drug and Microcrystalline cellulose as functional excipient) was integrated with the computer algorithm using Quantum Geographic Information System (QGIS) program (typically used for geographic information system application), Python, and SciPy. A 60:40 (drug: MCC) mixture was prepared experimentally, and obtained ratio from automatic image analysis is 58.5: 41.5, which can be improved further. The efficacy of current research is critical to creating a ground truth microscopic training image dataset for quick and auto determination of particle distribution using deep learning.

This document is currently not available here.