Other types of numerical features were also explored for their ability to describe protein localization patterns. A set of descriptive features that are fundamentally different from the Zernike moments, the texture features described by Haralick [28], were investigated next. These features were selected because they can be made invariant to translations and rotations, and because they describe more intuitive aspects of the images (e.g. coarse versus smooth, directionality of the pattern, image complexity, etc.) using statistics of the gray-level co-occurrence matrix for each image.
Since the BPNN classifier proved more effective than the
classification tree when using the Zernike features, it was used with
the texture features. This time the network had only 13 inputs but
still had 20 hidden nodes and 5 output nodes, all fully connected.
Training was carried out as before and the results are shown in Table
2.6. The average performance of this feature
set/classifier combination,
- corresponding to a predicted
accuracy of 99.6% for majority rule on 10 images - is very close to
that of the Zernike moment/BPNN approach. Note that this performance
is accomplished with far fewer features describing each image, 13
texture features versus 49 Zernike moments.
| True | Output of the BPNN | ||||
| Classification | Giantin | Hoechst | LAMP2 | NOP4 | Tubulin |
| Giantin | 97% | 0% | 3% | 0% | 0% |
| Hoechst | 3% | 93% | 3% | 0% | 0% |
| LAMP2 | 8% | 0% | 82% | 8% | 2% |
| NOP4 | 0% | 0% | 13% | 88% | 0% |
| Tubulin | 4% | 0% | 8% | 8% | 81% |
95% confidence interval). Performance
across the 8 test sets ranged from 70-98%. Average performance on
the training data was
.