Although research on estimating cancer genotype using radiomic features has progressed, radioproteomics research on estimating proteins has not progressed. The purpose of this study is to construct a method to discriminate positive and negative hormone receptors using radiomic features. We selected contrast-enhanced T₁-weighted fat suppression images of 49 patients (37 hormone receptor positive, 12 negative) from the public database TCGA-BRCA. T₂-weighted fat suppression images of 122 patients (72 hormone receptor positive, 50 negative) were also selected from the different public database I-SPY1. The slice with the largest tumor diameter was selected and the tumor region was manually marked, and 275 radiomics features were extracted from the tumor region. Logistic regression with radiomic features selected by Lasso was employed for discriminating between positive and negative hormone receptors. As the result of ROC analysis, AUC was 0.77 when contrast-enhanced T₁-weighted fat suppression images were used, and AUC was 0.62 when T₂-weighted fat suppression images were used. Therefore, the results suggest that contrast-enhanced T₁-weighted fat suppression images are more useful in discriminating positive and negative hormone receptors. In conclusion, non-invasive, low-cost imaging test can be used in radioproteomics studies to select patients for whom hormonal therapy is effective.

In radiographic examination, in case of the patient’s body movement or incomplete breath-holding, radiography was retaken again. Therefore, we investigated in U-net, Cycle-GAN, and UNIT, and developed a method for remove blur on radiographs with higher accuracy. The database used in this study consists of 39 cases of pair of radiography (blurred image and reference image) at the Eastern Chiba Medical Center, in which the radiography was retaken due to body movement or incomplete breath-holding. Next, for data augmentation, the number of training images was increased by rotation and inversion, the blurred image was input to U-net, Cycle-GAN and UNIT to generate no-blurred image. Comparing with the reference image, four radiological technologists performed visual evaluation of the test images on 5-point scale. Blurred score, retention of anatomy, contrast change on visual evaluation in UNIT was higher than this in U-net and Cycle-GAN, and there was a statistically significant difference (P<0.01). In the future, it will be necessary to increase the number of cases and further improve the accuracy.

To visualize widening of the intervertebral discspace (IDS), X-rays areirradiated tangentially on the articular surface. However, accurately estimatingthe deflection angle of the x-ray tube is difficult. Therefore, we classifiedthe standing lateral lumbar spine images into 5 types by a Deep ConvolutionalNeural Network, determined the deflection angle of the X-ray tube for moreefficient visualization of the extension of the IDS in the standing frontallumbar spine images. In the lumbar lateral images of 500 patients, the x-raytube deflection angle was measured manually for each intervertebral space, and 1723 regions cut out into 256×256 regions per vertebra wereused．Data augmentation increased the data to 3795 regions. The accuracy, precision and recall were 83.0 %, 84.1% and 83%, respectively, and the f-value was 83.3%, resulting in a relatively highclassification accuracy. Many patientsare already having lower back pain, requiring them to shift from the lateral tofrontal body positions swiftly. For this reason, using deep learning wouldenable taking the IDS measurement in a short time, thereby reducing burden onthe patient and improving the imaging flow.