A deep image classification model based on prior feature knowledge embedding and application in medical diagnosis.

Aiming at the problem of image classification with insignificant morphological structural features, strong target correlation, and low signal-to-noise ratio, combined with prior feature knowledge embedding, a deep learning method based on ResNet and Radial Basis Probabilistic Neural Network (RBPNN) is proposed model. Taking ResNet50 as a visual modeling network, it uses feature pyramid and self-attention mechanism to extract appearance and semantic features of images at multiple scales, and associate and enhance local and global features. Taking into account the diversity of category features, channel cosine similarity attention and dynamic C-means clustering algorithms are used to select representative sample features in different category of sample subsets to implicitly express prior category feature knowledge, and use them as the kernel centers of radial basis probability neurons (RBPN) to realize the embedding of diverse prior feature knowledge. In the RBPNN pattern aggregation layer, the outputs of RBPN are selectively summed according to the category of the kernel center, that is, the subcategory features are combined into category features, and finally the image classification is implemented based on Softmax. The functional module of the proposed method is designed specifically for image characteristics, which can highlight the significance of local and structural features of the image, form a non-convex decision-making area, and reduce the requirements for the completeness of the sample set. Applying the proposed method to medical image classification, experiments were conducted based on the brain tumor MRI image classification public dataset and the actual cardiac ultrasound image dataset, and the accuracy rate reached 85.82% and 83.92% respectively. Compared with the three mainstream image classification models, the performance indicators of this method have been significantly improved.

[Spatial Heterogeneity of Soil Respiration in a Planted Larch Forest in Shanxi Plateau].

Based on the data from a planted larch forest in Panquangou Natural Reserve of Shanxi Province, at three sampling scales (4, 2, and 1 m, respectively), soil respiration (Rs) and its affecting factors including soil temperature at 5 cm (T5), 10 cm (T10), and 15 cm (T15) depths, soil water content (Ws), litter mass (Lw), litter moisture (Lm), soil total carbon (C), and soil total nitrogen ( N) were determined. The spatial heterogeneities of Rs and the environmental factors were further analyzed and their intrinsic correlations were established. The results of traditional statistics showed that the spatial variations of Rs and the all measured factors were in the middle range; Rs were highly significantly positively correlated with T10, T15, and N (P < 0.01); significantly positively correlated with Lm (P < 0.05); highly significantly negatively correlated with C/N ratio (P < 0.01); and not significantly correlated with T5, Ws, Lw and C (P > 0.05). Multiple stepwise regression analysis indicated that the four factors of Lm, T10, N, and Ws together accounted for 36% of Rs heterogeneity. The results of geo-statistical analysis demonstrated that Rs was in a medium spatial autocorrelation; random and structural factors accounted for 39.5% and 60.5% of Rs heterogeneity, respectively. And the factors such as climate, landform, and soil played a leading role. The results also illustrated that the ranges for soil factors were different and the range for both Rs and T10 was 25 meters. The fractal dimension of the soil index was in the following order: Lw and C/N ratio (1.95) > N (1.91) > C (1.89) > Rs (1.78) > Lm (1.77 ) > Ws (1.69) > T10 (1.42). The spatial distribution of Rs was in consistent agreement with those of T10, Lm, C, and N; but different with those of Ws and C/N ratio. With a fixed cofidence level and certain estimated accuracy, the required sampling number of each item differed, corresponding to its spatial variation degree.