Objective: Radiomics and deep transfer learning are two popular technologies used to develop computer-aided detection and diadnosis (CAD) schemes of medical images. This study aims to investigate and compare advantages and potential limitations of applying these two technologies in developing CAD schemes. Methods: A relatively large and diverse retrospective dataset including 3,000 digital mammograms is assembled in which 1,496 images depict malignant lesions and 1,054 images depict benign lesions. Two CAD schemes are developed to classify breast lesions. The first scheme is developed using four steps namely, applying an adaptive multi-layer topographic region growing algorithm to segment lesions, computing initial radiomics features, applying a principal component algorithm to generate an optimal feature vector, and building a support vector machine classifier. The second CAD scheme is built based a pre-trained residual net architecture (ResNet50) as a transfer learning model to classify breast lesions. Both CAD schemes are trained and tested using a 10-fold cross-validation method. Several score fusion methods are also investigated to classify breast lesions. CAD classification performance is evaluated by the area under ROC curve (AUC). Results: ResNet50 model-based CAD scheme yields AUC = 0.85±0.02, which is significantly higher than radiomics feature-based CAD scheme with AUC = 0.77±0.02 (p < 0.01). Additionally, fusion of classification scores generated by two CAD schemes does not further improve classification performance. Conclusion: This study indicates that using deep transfer learning is more efficient to develop CAD schemes and enables to yield higher lesion classification performance than CAD schemes developed using radiomics-based technology.

Deep convolutional neural networks (CNNs) have been widely used in various medical imaging tasks. However, due to the intrinsic locality of convolution operation, CNNs generally cannot model long-range dependencies well, which are important for accurately identifying breast cancer from unregistered multi-view mammograms. This motivates us to leverage the architecture of Vision Transformers to capture long-range relationships of multiple mammograms from the same patient. For this purpose, we employ local Transformer blocks to separately learn patch relationships within a specific view (CC/MLO) of one side (right/left) of mammogram. The outputs from different views and sides are concatenated and fed into global Transformer blocks, to jointly learn patch relationships between two different views of the left and right breasts. To evaluate the proposed model, we retrospectively assembled a dataset involving 949 sets of CC and MLO view mammograms, which include 470 malignant cases and 479 normal or benign cases. We trained and evaluated the model using a five-fold cross-validation method. Without any arduous preprocessing steps (e.g., optimal window cropping, chest wall or pectoral muscle removal, two-view image registration, etc.), our two-view Transformer-based model achieves lesion classification accuracy of 77.0% and area under ROC curve (AUC = 0.814), outperforming state-of-the-art multi-view CNNs by 3.1% and 3%, respectively. Meanwhile, the new two-view model improves mammographic case classification accuracies of two single-view models by 7.4% (CC) and 4.5% (MLO), respectively. The promising results unveil the great potential of using Transformers to develop high-performing computer-aided diagnosis (CADx) schemes of mammography.

Assessment of the percentage of disease infected lung volume using computed tomography (CT) images can play an important role to detect lung diseases and predict disease severity. However, manual segmentation of disease infected regions from many CT image slices is tedious and not feasible in clinical practice. To help solve this clinical challenge, this study aims to investigate a new strategy to automatically segment disease infected regions and predict disease severity. We employed a public dataset acquired from 20 COVID-19 patients, which includes manually annotated lung and infections masks, to train a new ensembled deep learning (DL) model that combines the five customized residual attention UNet models to segment disease infected regions followed by a Feature Pyramid Network (FPN) model to classify severity stage of COVID-19 infection. To test potentially clinical utility of new model, we first gathered and processed another set of CT images acquired from 80 Covid-19 patients. Next, we asked two chest radiologists to read CT images of each patient and report the estimated percentage of infected lung volume and disease severity level. Additionally, we asked radiologists to rate acceptance of DL model-generated segmentation results using a 5-scale rating method. Data analysis results show that agreement between disease severity classification is &gt;90% in 45 testing cases. Furthermore, &gt;73% of cases received the high rating score from two radiologists (scored more than 4). This study demonstrates feasibility of developing a new DL-model to efficiently provide quantitative assessment of disease severity based on the automated segmentation of the disease infected regions to support improving efficacy of radiologists in disease diagnosis.

MADS-box genes play a crucial role in fruit ripening, yet limited research has been conducted on mango. Based on the conserved domains of this gene family, 84 MADS-box genes were identified in the mango genome, including 22 type I and 62 type II MADS-box genes. Gene duplication analysis revealed that both tandem duplication and segmental replication significantly contributed to the expansion of MADS-box genes in the mango genome, with purifying selection playing a vital role in the segmental duplication events within the MiMADS gene family. Cis-acting element analysis demonstrated that most MiMADS genes were hormonally regulated and participated in the growth, development, and stress resistance of mango fruit. Moreover, through expression pattern analysis and phylogenetic tree construction, we identified six MiMADS genes belonging to the SEP1 subfamily and two belonging to the AG subfamily as potential candidates involved in mango ripening regulation. Notably, Mi08g17750 and Mi04g18430 from the SEP1 subfamily were identified as key regulators inhibiting mango fruit maturation; their interaction network was also analyzed. These findings provide a foundation for further investigation into the regulatory mechanisms underlying mango ripening.

Based on screening in computational biology and biological in vitro assays, five natural products isolated from extracts of the herbal medicine toad skin, such as cinobufagin (CBFi), bufalin (BFi), arenobufagin (ABFi), telocinobufagin (TBFi), bufotalin (BFTi), were subjected to molecular docking calculations with the use of SARS-CoV-2 main protease (PDB 6LU7 and 7BTF) and top-scoring ligand-receptor complexes were obtained. The results showed that the binding energy of ABFi to the 3CL protein was -17.044kcal/mol, which was higher than CBFi and TBFi. However, the binding energy of ABFi to the RdRp protease was -23.250 kcal/mol, which was much lower than that of CBFi and TBFi, EVEN lower than that of ABFi to the 3CL protein. ABFi also has polar interactions with amino acids such as Glu811, Ser814, Ser681 and Thr680 of RdRp enzyme. The results revealed that ABFi had a moderate inhibitory effect on the cell proliferation of SARS-CoV-2 in vitro, with an inhibition rate of 61.12%, even weaker than Remdesivir. This new discovery provides us with new ideas for in-depth studies on the development of natural products with this class of structural generalizations as inhibitors of SARS-CoV-2, and provides an experimental basis for the next step of mechanistic studies.

Intercropping of maize (Zea mays L.) and peanuts (Arachis hypogaea L.) (M||P) significantly enhances crop yield. In a long-term M||P field experiment with two P fertilizer levels, we examined how long-term M||P affects topsoil aggregate fractions and stability, organic carbon (SOC), available phosphorus (AP), and total phosphorus (TP) in each aggregate fraction, along with crop yields. Compared to their respective monocultures, long-term M||P substantially increased the proportion of topsoil mechanical macroaggregates (7.6–16.3%) and water-stable macroaggregates (>1 mm) (13.8–36.1%), while reducing the unstable aggregate index (ELT) and the percentage of aggregation destruction (PAD). M||P significantly boosted the concentration (12.9–39.9%) and contribution rate (4.1–47.9%) of SOC in macroaggregates compared to single crops. Moreover, the concentration of TP in macroaggregates (>1 mm) and AP in each aggregate fraction of M||P exceeded that of the respective single crops (P<0.05). Furthermore, M||P significantly increased the Ca2-P, Ca8-P, Al-P, and Fe-P concentrations of intercropped maize (IM) and the Ca8-P, O-P, and Ca10-P concentrations of intercropped peanuts (IP). The land equivalent ratio (LER) of M||P was higher than one, and M||P stubble improved the yield of subsequent winter wheat (Triticum aestivum L.) compared with sole-crop maize stubble. P application augmented the concentration of SOC, TP, and AP in macroaggregates, resulting in improved crop yields. In conclusion, our findings suggest that long-term M||P combined with P application, sustains farmland productivity in the North China Plain by increasing SOC and macroaggregate fractions, improving aggregate stability, and enhancing soil P availability.