ABSTRACT:To create image objects for subsequent classification in object-based image analysis, an optimal segmentation threshold (OST) is a pre-requisite for image segmentation. However, an OST is practically acquired by assessing and ranking an exhaustive segment data stack constructed after foregoing image segmentation. In this paper, we propose an iterative exploration method via recognition of the Euclidean distance 2 (ED2)-scale parameter (SP) pattern with the least five tiles of segment data stack in each cycle among the Potential Segmentation Error (PSE)-Number-of-Segments Ratio (NSR)-ED2 patterns along with SP. We conducted two experiments. In the first experiment, we validated the general italic U-shaped ED2-SP pattern by constructing exhaustive segment data stacks and corresponding segment data stacks. In the second experiment, we adopted the proposed iterative exploration method for OST selection based on the ED2-SP pattern with respect to five equal-spacing SPs in each cycle. The bottom of the pattern was persistently approached by constructing updated segment data stacks and corresponding segment data stacks with five dynamically adjusted tiles. Our results showed that the PSE-NSR-ED2 discrepancy measure system is advantageous to OST selection.

No abstract

Background To improve the accuracy of pneumoconiosis diagnosis, a computer-assisted method was developed. Methods Three CNNs (Resnet50, Resnet101, and DenseNet) were used for pneumoconiosis classification based on 1,250 chest X-ray images. Three double-blinded experienced and highly qualified physicians read the collected digital radiography images and classified them from category 0 to category III. The results of the three physicians in agreement were considered the relative gold standards. Subsequently, three CNNs were used to train and test these images and their performance was evaluated using multi-class classification metrics. We used kappa values and accuracy to evaluate the consistency and reliability of the optimal model with clinical typing. Results ResNet101 was the optimal model among the three CNNs. The AUC of ResNet101 was 1.0, 0.9, 0.89, and 0.94 for detecting pneumoconiosis categories 0, I, II, and III, respectively. The micro-average and macro-average mean AUC values were 0.93 and 0.94, respectively. The accuracy and Kappa values of ResNet101 were 0.72 and 0.7111 for quadruple classification and 0.98 and 0.955 for dichotomous classification, respectively, compared with the relative standard classification of the clinic. Conclusion The ResNet101 model performed relatively better in classifying pneumoconiosis than radiologists. The dichotomous classification displayed outstanding performance, thereby indicating the feasibility of deep learning techniques in pneumoconiosis screening.