Neural computing in pharmaceutical products and process development - Computer-Aided Applications in Pharmaceutical Technology - page 115

Modeling of the subdivision of scored tablets.
The amounts of the materials and the
compression force were used as input variables;
whereas parameters describing the compression
process (elastic recovery, lubrication, plasticity,
etc.) and post-compression parameters (tablet
hardness, true density and breaking into halves),
were investigated as outputs.

GRNN,
RBFNN,
MLP

Sovány
et al., 2009

Modeling of encapsulation and release of an
ionizable drug from polymer microspheres. Inputs
used for the network training were pH of the
external aqueous phase, polymer concentration,
and the initial drug load; whereas the outputs
were percentage of drug released, mean size
diameter, and yield of microspheres.

MLP Labouta

et al., 2009

Testing of the stability of hydrocortisone powder
for injections. Degradation products of
hydrocortisone sodium succinate were followed
during stress and formal stability studies.

DNN Solomun

et al., 2012

Formulation, characterization and optimization of modifi ed/extended/
controlled release preparations; in vitro - in vivo correlations
Design of controlled release formulations. Varying
formulation variables were used as inputs and
in vitro cumulative percentages of drug released
were used as ANN outputs.

MLP Chen et al.,

1999

Optimization of diclofenac sodium sustained
release matrix tablets. Trained model was used
to predict release profi les and to optimize the
formulation composition.

MLP Zupancic

Bozic et al.,
1997

Design of extended release aspirin tablets.
The amount of Eudragit® RS PO polymer and
compression pressure were selected as inputs,
whereas in vitro dissolution profi les, release
order, and release constant from the Peppas
equation were selected as output parameters.

GRNN Ibric et al.,

2002

￿

￿ ￿

Prediction of drug dissolution profi les. Inputs for
the network training were the matrix forming
agents' ratio, the time point of the measurement
of percentage of drugs dissolved, and the
difference between the release rates of the
preceding two time points of the predicted profi le.
In vitro dissolution profi les were used as network
outputs.

MLP Peh et al.,

2000

( Continued )