Japanese journal of radiology
-
Deep learning with a convolutional neural network (CNN) is gaining attention recently for its high performance in image recognition. Images themselves can be utilized in a learning process with this technique, and feature extraction in advance of the learning process is not required. Important features can be automatically learned. ⋯ This article illustrates basic technical knowledge regarding deep learning with CNNs along the actual course (collecting data, implementing CNNs, and training and testing phases). Pitfalls regarding this technique and how to manage them are also illustrated. We also described some advanced topics of deep learning, results of recent clinical studies, and the future directions of clinical application of deep learning techniques.
-
Smoking-related lung abnormalities are now an increasing public health concern. According to the findings of large-cohort studies, approximately 8% of smokers have interstitial lung abnormalities, which are associated with a relatively high risk of all-cause mortality. We reviewed the radiological and pathological findings of smoking-related interstitial lung diseases, such as respiratory bronchiolitis-interstitial lung disease, desquamative interstitial pneumonia, and airspace enlargement with fibrosis. ⋯ Therefore, diagnosis should be performed not on the basis of a single radiological finding, but in a comprehensive manner, by including clinical symptoms and disease behavior. Among interstitial abnormalities in smokers, the usual interstitial pneumonia (UIP) pattern is correlated with a worse prognosis than others. Basal-predominant subpleural reticulation is a clue for accurate diagnosis of UIP, which can be achieved by computer-aided quantitative analysis.
-
To distinguish between adenocarcinoma in situ (AIS)-minimally invasive adenocarcinoma (MIA) and invasive adenocarcinoma (IAC) showing pure or part-solid ground-glass nodules (GGNs) by high-resolution computed tomography (HRCT) texture analysis. ⋯ The 90th percentile CT numbers and entropy can accurately distinguish AIS-MIA from IAC.
-
To investigate the impact of three-dimensional (3D) T2-weighted turbo spin-echo imaging (TSE-T2WI) with tissue-specific variable refocusing flip angle (TS-VRFA) on image quality and prostate cancer (PCa) detection and extraprostatic extension (EPE) evaluation compared to 2D TSE-T2WI and conventional 3D TSE-T2WI with volume isotropic TSE acquisition (VISTA). ⋯ 3D T2 WI using TS-VRFA could potentially replace multiplane 2D T2 WI for prostate cancer diagnosis with better image quality than VISTA.
-
This study aimed to compare the tumor volume definition using conventional magnetic resonance (MR) and 11C-methionine positron emission tomography (MET/PET) images in the differentiation of the pre-operative glioma grade by using whole-tumor histogram analysis of normalized cerebral blood volume (nCBV) maps. ⋯ Whole-tumor histogram analysis of CBV map provides more valuable histogram parameters and increases diagnostic accuracy in the differentiation of pre-operative cerebral gliomas when the tumor volume is derived from MET/PET images.