Compared to computed tomography (CT), magnetic resonance imaging (MRI) is more sensitive to acute ischemic stroke lesion. However, MRI is time-consuming, expensive, and susceptible to interference from metal implants. Generating MRI images from CT images can address the limitations of MRI. The key problem in the process is obtaining lesion information from CT. In this study, we propose a cross-modal image generation algorithm from CT to MRI for acute ischemic stroke by combining radiomics with generative adversarial networks. First, the lesion candidate region was obtained using radiomics, the radiomic features of the region were extracted, and the feature with the largest information gain was selected and visualized as a feature map. Then, the concatenation of the extracted feature map and the CT image was input in the generator. We added a residual module after the downsampling of the generator, following the general shape of U-Net, which can deepen the network without causing degradation problems. In addition, we introduced the lesion feature similarity loss function to focus the model on the similarity of the lesion. Through the subjective judgment of two experienced radiologists and using evaluation metrics, the results showed that the generated MRI images were very similar to the real MRI images. Moreover, the locations of the lesions were correct, and the shapes of lesions were similar to those of the real lesions, which can help doctors with timely diagnosis and treatment.

This study aims at analyzing the separability of acute cerebral infarction lesions which were invisible in CT. 38 patients, who were diagnosed with acute cerebral infarction and performed both CT and MRI, and 18 patients, who had no positive finding in either CT or MRI, were enrolled. Comparative studies were performed on lesion and symmetrical regions, normal brain and symmetrical regions, lesion, and normal brain regions. MRI was reconstructed and affine transformed to obtain accurate lesion position of CT. Radiomic features and information gain were introduced to capture efficient features. Finally, 10 classifiers were established with selected features to evaluate the effectiveness of analysis. 1301 radiomic features were extracted from candidate regions after registration. For lesion and their symmetrical regions, there were 280 features with information gain greater than 0.1 and 2 features with information gain greater than 0.3. The average classification accuracy was 0.6467, and the best classification accuracy was 0.7748. For normal brain and their symmetrical regions, there were 176 features with information gain greater than 0.1, 1 feature with information gain greater than 0.2. The average classification accuracy was 0.5414, and the best classification accuracy was 0.6782. For normal brain and lesions, there were 501 features with information gain greater than 0.1 and 1 feature with information gain greater than 0.5. The average classification accuracy was 0.7480, and the best classification accuracy was 0.8694. In conclusion, the study captured significant features correlated with acute cerebral infarction and confirmed the separability of acute lesions in CT, which established foundation for further artificial intelligence‐assisted CT diagnosis.

Data‐hunger is a persistent challenge in machine learning, particularly in the field of image processing based on convolutional neural networks (CNNs). This study systematically investigates the factors contributing to data‐hunger in machine‐learning‐based image‐processing algorithms. The results revealed that the proliferation of model parameters, the lack of interpretability, and the complexity of model structure are significant factors influencing data‐hunger. Based on these findings, this paper introduces a novel semi‐white‐box neural network model construction strategy. This approach effectively reduces the number of model parameters while enhancing the interpretability of model components. It accomplishes this by constraining uninterpretable processes within the model and leveraging prior knowledge of image processing for model. Rather than relying on a single all‐in‐one model, a semi‐white‐box model is composed of multiple smaller models, each responsible for extracting fundamental semantic features. The final output is derived from these features and prior knowledge. The proposed strategy holds the potential to substantially decrease data requirements under specific data source conditions while improving the interpretability of model components. Validation experiments are conducted on well‐established datasets, including MNIST, Fashion MNIST, CIFAR, and generated data. The results demonstrate the superiority of the semi‐white‐box strategy over the traditional all‐in‐one approach in terms of accuracy when trained with equivalent data volumes. Impressively, on the tested datasets, a simplified semi‐white‐box model achieves performance close to that of ResNet while utilizing a small number of parameters. Furthermore, the semi‐white‐box strategy offers improved interpretability and parameter reusability features that are challenging to achieve with the all‐in‐one approach. In conclusion, this paper contributes to mitigating data‐hunger challenges in machine‐learning‐based image processing through the introduction of a novel semi‐white‐box model construction strategy, backed by empirical evidence of its effectiveness.

Remote sensing image change detection (RSCD) aims to extract high-level semantic feature (HLSF) to identify the changed areas (CAs) between dual-temporal images (DTIs). However, the diversity in the CA shape and size poses certain challenge to the CD task. Besides, different illumination conditions in the same scene of the DTI further increase the CD difficulty. In response to these above issues, this paper proposes a multi-scale feature (MSF) fusion CD network-WSMsFNet, which fully utilizes the local and global information of multi-scale features to achieve comprehensive representation of the change scene; In addition, the network improves the feature extraction ability of each module through the whole process supervision loss function. Firstly, the network hierarchically extracts different scale information of the two temporal RS; Then, special information enhancement and fusion modules are constructed for various feature layers (i.e., the same level, adjacent level, and global features), aiming to enhance the local feature representation ability and contextual information relevance of the deep network; Finally, the whole-process loss function is set to supervise the intermediate layer learning, which can effectively enhance the feature representation ability and guide feature extraction direction of each module. Experiments have shown that the WSMsFNet has achieved significant results in both qualitative and quantitative indicators.