In Japan, breast MRI is widespread as a diagnostic modality following mammography and ultrasound. The main purpose of performing breast MRI is to determine the extent of carcinoma, especially in ductal carcinoma in situ (DCIS) .For the diagnosis of DCIS, a morphological evaluation is more useful than the time-intensity curve in contrast-enhanced MRI. Recently, neoadjuvant chemotherapy for breast cancer has increased and MRI is considered to be the best method to evaluate its effect. MR spectroscopy and MRI-guided biopsy are also promising methods, however, there are many problems related to their clinical use.
The number of patients with osteoporosis is increasing every year in Japan. The bone-mineral-density (BMD) and spinal curvature are two important factors related to the fractures and osteoporosis. Recently, multi-detector-row CT images are widely used in clinical medicine and may have the potential for osteoporosis diagnosis. The development of a computer-aided diagnosis (CAD) system that can support the osteoporosis diagnosis is now required. This study proposes a method that can measure the distributions of BMD in each of vertebral trabecular bones and angles of spinal curvatures automatically on torso CT images. This method extracts the regions of the vertebral bones firstly and then generates a sagittal plane of the human spine. Finally, 4 corner points of each vertebral bone region are identified and used for measuring the BMD distributions and spinal curvatures. This method was applied to 20 CT cases and the results were compared with the gold standard that were generated by a medical expert. The errors of the average value of BMD have the mean value of 6.93 mg/cm3 (std.dev.: 6.82) . The angles have the mean value of 4.62 degrees (std. dev.: 3.23) in thoracic region and 4.40 degrees (std. dev.: 5.08) in abdomen. These results showed a good performance of our purposed method and possibility for supporting the CAD of osteoporosis diagnosis.
The detection of asymptomatic lacunar infarcts on magnetic resonance (MR) images is important because their presence indicates an increased risk of severe cerebral infarction. However, accurate identification of lacunar infarcts on MR images is often hard for radiologists because of the difficulty in distinguishing lacunar infarcts and enlarged Virchow-Robin spaces. Therefore, we developed a computer-aided diagnosis (CAD) scheme for the classification of lacunar infarcts and enlarged Virchow-Robin spaces. Our database consisted of T1-and T2-weighted images obtained from 52 patients, which included 89 lacunar infarcts and 20 enlarged Virchow-Robin spaces. The locations of lacunar infarcts and enlarged Virchow-Robin spaces were determined by experienced neuroradiologists. We first enhanced the lesions in T2-weighted image by using the white top-hat transformation. A gray-level thresholding was then applied to the enhanced image for the segmentation of lesions. From the segmented lesions, we determined image features, such as size, shape, location, and signal intensities in T1-and T2-weighted images. A neural network was then employed for distinguishing between lacunar infarcts and enlarged Virchow-Robin spaces. Our computerized method was evaluated by using a leave-one-out method. The result indicated that the area under the ROC curve was 0.893. Therefore, our CAD scheme would be useful in assisting radiologists for distinguishing between lacunar infarcts and enlarged Virchow-Robin spaces in MR images.
We evaluated the R2* (iron content) and fat fraction of the liver tissue simultaneously using modulus and real multiple gradient-echo (MRM-GRE) sequence. On a 1.5-T MRI, modulus images of 7-9 gradient-echoes were obtained by MRM-GRE sequence at a time. A real part of the first echo image was also reconstructed to differentiate below and above the 50 percent fat fraction. The fat fraction and R2* were obtained from the parameters of a theoretically fitted formula with each echo signal. R2* and fat fraction were measured with MRM-GRE in the phantom and the liver in normal volunteers (n=6) and patients with fatty liver (n=4) . MRI-derived fat fraction of the phantom was in good agreement with the actual value, and R2* of the phantom showed a strongly positive correlation with the actual iron content. MRI-derived fat fraction in fatty liver was significantly higher than that in the normal volunteers. However, no significant difference in R2* was found between fatty liver and normal volunteers. These results show that the MRM-GRE enables to differentiate the causes of signal reduction whether increasing R2*or increasing fat fraction. The MRM-GRE method makes it possible to simply and accurately assess the fat content and the iron content.
Pulmonary ventilation and blood flow are reflected in dynamic chest radiographs as changes in X-ray translucency, i.e., pixel values. Thus, relative local pulmonary function can be evaluated based on changes in pixel value. The purpose of this study was to investigate the feasibility of ventilation-perfusion evaluation (V/Q study) based on changes in pixel value in dynamic chest radiographs. Sequential chest radiographs of a patient with ventilation-perfusion mismatch were obtained with a dynamic FPD system. The changes in pixel value resulting from respiration and blood flow were measured in respiratory phase and breath-holding phase, respectively, and the radio was calculated in each local area. The results were compared to distribution of radioactivity counts and V/Q. In the results, abnormalities were appeared as a reduction of changes in pixel values, and a correlation was observed between the distribution of changes in pixel value and those of radioactivity counts (Ventilation; r=0.78, Perfusion; r=0.77) . Ventilation-perfusion mismatch was also indicated as mismatch of changes in pixel value, and a correlation with V/Q calculated by radioactivity counts (r=0.78) . The present method is potentially useful for V/Q study as an additional examination in conventional chest radiography.
The automated segmentation of the prostate region in CT images is required by the computer-aided diagnosis and therapeutic radiology. The purpose of this study is to develop an automated scheme for estimating the location (center point) of the prostate gland in torso X-ray CT images. The proposed scheme consists of four processing steps. At first, a prostate gland centroid is extracted manually, and pelvis landmarks are extracted automatically on training dataset. Second, based on the pelvis landmarks, each pelvis shape is normalized. Third, a prostate gland centroid model is constructed by the principal component analysis. Finally, a prostate gland model is applied to a target case, and the center point of prostate gland is estimated. The proposed scheme was applied to 20 CT cases. We found the proposed scheme could estimate the location of the prostate gland with the mean error (3-D distance from the grand truth) of 2.6mm. This scheme was used for prostate gland segmentation using a simple sphere model. From experimental results, the effect of the proposed scheme for estimating the location of the prostate gland was confirmed.