Neural computing in pharmaceutical products and process development - Computer-Aided Applications in Pharmaceutical Technology - page 114

Review of various applications of the ANN
methodology ( Continued)

Table 5.2

Application ANN used Reference

Optimization of the tablet coating formulation.
Inputs for the network training were molecular
weight of the polymer used for coating, pigment
type, pigment particle size and size distribution,
pigment concentration, and fi lm thickness.
Network outputs were crack velocity and fi lm
opacity.

MLP Plumb

et al., 2002

Determination of factors controlling the
particle size in nanoemulsions. Inputs for the
neural network were both formulation and
processing factors: percentage of ethanol in
the formulation, the amount of budesonide
dissolved in the preparation, saline normality,
total energy applied, and the rate of energy
applied. The single output was particle
(droplet) size.

MLP Amani

et al., 2008

Optimization of the freeze-dried, mucoadhesive
system for vaginal delivery of the drug. Network
inputs were hydroxypropyl-methylcellulose:
polyvinylpyrrolidone ratio, mucoadhesive
concentration, and the type of delivery
system; whereas outputs were cumulative
release of the drug, mucoadhesive force and
zero-rate viscosity.

MLP Woolfson

et al., 2010

Characterization of the powder fl ow rate relevant
to mini-tableting. Neural network inputs were bulk
density, bulk/tapped density ratio, orifi ce
diameter, particle diameter, AR, roundness,
convexity, particle density, and die thickness. The
single output was the fl ow rate.

MLP Kachrimanis

et al., 2005

￿

￿ ￿

Development and formulation of push-pull
osmotic pump tablets. Drug release was
predicted using a neural network model.

MLP Zhang

et al., 2011

Characterization of the tablet database
containing several active ingredients.
Formulation factors were used to predict
tablet properties.

EANN Takagaki

et al., 2010