Table 11. Comparison of the results of related method and proposed method for classifying multiple types of lung nodules.
| Nodule’s types | Chen et al. (52) | Ni et al. (4) | Ours (M2) | |||||
|---|---|---|---|---|---|---|---|---|
| ACC | ADE | ACC | ADE | ACC | ADE | |||
| Sphericity | – | 0.8600 | 0.7121 | 0.7000 | 0.6669 | 0.6662 | ||
| Lobulation | – | 0.8000 | 0.9566 | 0.5000 | 0.6897 | 0.6205 | ||
| Spiculation | – | 0.6400 | 0.9431 | 0.5000 | 0.6869 | 0.6261 | ||
| Texture | – | 0.1800 | 0.7888 | 0.4800 | 0.8762 | 0.2473 | ||
| Margin | – | 0.9200 | 0.8111 | 0.7400 | 0.7826 | 0.4346 | ||
| Calcification | – | 0.8700 | 0.9221 | 0.3100 | 0.9642 | 0.0714 | ||
| Malignancy | – | 0.8700 | 0.8661 | 0.5900 | 0.8168 | 0.3662 | ||
The ADE is a metric that quantifies the disparity between the model’s predicted value and the actual value. A smaller ADE indicates the model’s superior predictive performance. In the classification of lobulation, spiculation, texture, and margin of nodules, this study performed a more detailed binary classification of nodules, while other studies were more general. To facilitate an approximate comparison, in this table, we averaged the metric results of the sub-tasks for each of these four classification tasks. ACC, accuracy; ADE, absolute distance error.