A 3 × 3 factorial feeding trial was conducted to evaluate the production response of juvenile northern snakehead fish (Channa argus). Nine diets containing 3 protein levels (45%, 48% and 51%) and 3 lipid levels (9%, 12% and 15%) were formulated and fed to triplicate groups of juvenile northern snakehead (15.78 ± 0.09 g/fish) for 8 weeks. The formulated diets were named as P45L9, P45L12, P45L15, P48L9, P48L12, P48L15, P51L9, P51L12 and P51L15 (P-Protein, L-Lipid), respectively. Fish fed diets with the lowest protein and lipid combination (P45L9) had the lowest growth performance. Weight gains (WG) of fish fed the 4 diets P48L12, P48L15, P51L9, and P51L12 were not significantly different (P > 0.05), but significantly higher (P < 0.05) than those of fish fed the other diets. Fish fed diets P48L12 and P48L15 had significantly lower (P < 0.05) feed conversion ratios (FCR) than the rest of the treatments. Protein retentions (PR) among fish fed the diets P45L12, P45L15, P48L12, P48L15, P51L9, and P51L12 were similar and significantly higher (P < 0.05) than those of fish fed the remaining diets. Protein sparing effect was observed in the treatments when fish was fed diets containing 45% or 48% dietary protein levels with dietary lipid increased from 9% to 12%. Fish fed diets with 9% lipid tended to have lower viscerosomatic index (VSI), hepatosomatic index (HSI), and whole-body lipid. Increasing dietary protein level significantly increased (P < 0.05) liver moisture and lipid while dietary lipid level increased liver lipid. Intestinal lipase activity increased significantly (P < 0.05) with increasing dietary lipid and protein levels while intestinal α-amylase and protease activities were not significantly influenced (P > 0.05) by dietary treatments. Based on these results, the diet containing 48% protein with either 12% or 15% lipid is the optimal for supporting growth and feed utilization of juvenile northern snakehead under the current testing conditions.

Six test diets with protein levels varying from 250 to 500 g kg )1 were fed to six triplicate groups of summerling (initial weight: 1.56 g) and seven test diets with protein levels varying from 200 to 500 g kg )1 were fed to seven triplicate groups of winterling (initial weight: 9.49 g) for 8 weeks.Weight gain (WG) and feed efficiency (FE) of summerling significantly increased with increasing dietary protein levels from 250 to 350 g kg )1 and slightly declined, but without statistical significance at a dietary protein level of 400 g kg )1 , then further significantly decreased with increasing protein levels to 450 and 500 g kg )1 ; WG of winterling increased significantly with increasing dietary protein levels from 200 to 300 g kg )1 (P < 0.05), and above this level, WG had a tendency to decrease with increasing dietary protein levels. Winterling fed diets with 300 and 400 g kg )1 of dietary protein had significantly higher FE than those fed other diets. WG data analysis by quadratic regressions showed that the optimum dietary protein levels required for the maximum growth of summerling and winterling were 374 and 355 g kg )1 of dry diet respectively. Protein efficiency ratio of both summerling and winterling negatively correlated with levels of dietary protein. The whole body moisture, protein, lipid and ash of summerling after being fed various test diets for 8 weeks were significantly different among treatments (P < 0.05). The whole body moisture and fat of winterling were also significantly affected by dietary protein levels (P < 0.05), while the whole body protein and ash of winterling were not (P > 0.05). KEY WORDS

Purpose. To study the application of medical image recognition technology based on backpropagation neural network (BPNN) in the effect of soybean meal replacing fish meal on intestinal microbial diversity of Channa argus and to evaluate the application value of this intelligent algorithm, Channa argus was fed with different contents of soybean meal instead of fish meal. Methods. After intestinal samples were collected and bacteria were isolated, microscopic imaging was performed, and the images were classified and identified. BPNN was constructed to perform denoising, smoothing, and segmentation. Results. After BPNN processing, the bacteria were completely separated from the original image background, and the bacteria was in the closed state, which was beneficial to feature extraction and species recognition. If there were 2 hidden layer nodes, the segmentation accuracy of bacterial microscopic images was the highest, up to 97.3%. With the replacement ratio of fish meal increased, the species of intestinal microbiome gradually enriched, and the relative abundance of intestinal microbiome was higher after fish meal was completely replaced by soybean meal (replacement). The intestinal microbial enzyme activities were affected by different fish meal and soybean meal contents in the diet. The glutamate transaminase and adenosine deaminase activities were increased after the replacement and were higher than those before the replacement, with statistically significant differences ( P < 0.05 ). Conclusion. Replacement of fish meal with soybean meal has a significant effect on the intestinal flora diversity of Channa argus, and there is a close relationship between them. The image recognition technology based on BPNN has high recognition rate and segmentation accuracy for microbiological microscopic images.