Search

Purpose Increased use of deep convolutional neural networks (DCNNs) in medical imaging diagnosis requires determinate evaluation of diagnostic performance. We performed the fundamental investigation of diagnostic performance of DCNNs using the detection task of brain metastasis. Methods We retrospectively investigated AlexNet and GoogLeNet using 3117 positive and 37961 negative MRI images with and without metastasis regarding (1) diagnostic biases, (2) the optimal K number of K-fold cross validations (K-CVs), (3) the optimal positive versus negative image ratio, (4) the accuracy improvement curves, (5) the accuracy range prediction by the bootstrap method, and (6) metastatic lesion detection by regions with CNNs (R-CNNs). Results Respectively, AlexNet and GoogLeNet had (1) 50 ± 4.6% and 50 ± 4.9% of the maximal mean ± 95% confidence intervals (95% CIs) measured with equal-sized negative versus negative image datasets and positive versus positive image datasets, (2) no less than 10 and 4 of K number in K-CVs fell within the respective maximum biases of 4.6% or 4.9%, (3) 74% of the highest accuracy with equal positive versus negative image ratio dataset and 91% of that with four times of negative-to-positive image ratio dataset, (4) the accuracy improvement curves increasing from 69% to 74% and 73% to 88% as positive versus negative pairs of the training images increased from 500 to 2495, (5) at least nine and six out of 10-CV result sets essential to predict the accuracy ranges by the bootstrap method, and (6) 50% and 45% of metastatic lesion detection accuracies by R-CNNs. Conclusions Our research presented methodological fundamentals to evaluate diagnostic features in the visual recognition of DCNNs. Our series will help to conduct the accuracy investigation of computer diagnosis in medical imaging.

Purpose To clarify the accuracy of vertebral puncture of the vertebral tertile area needling (VETERAN) method puncturing the pedicle superimposed on one-third of the width between the lateral vertebral line to the contralateral medial lamina line compared with Cathelin-needle-assisted puncture (CAP) method puncturing using the Cathelin needle as a guide in percutaneous vertebroplasty. Materials and methods 449 punctures by CAP method and 125 punctures by VETERAN method were enrolled. We compared the puncture accuracy of both methods. We estimated a vertebral estimated tilting ratio (VET-ratio) defined as ratio of the distance between the lateral vertebral line and the contralateral medial laminal line to the distance between the vertebral lateral line and the puncture point measured by computed tomography. We also estimated the procedural items and clinical outcomes. Results VETERAN method with 100% of punctures within safe zone (cortical breaches within 2 mm) had significantly higher accuracy than CAP method with 97.8% (p p p Conclusion VETERAN method reduced the occurrence of the inaccurate puncture, operative times, and exposure times. A VET-ratio with 36% or less is associated with a safe puncture using VETERAN method.

We performed CT-guided percutaneous uterine biopsy in three females for whom diagnosis by endometrial brushing cytology and endometrial biopsy had proved difficult. In each case, a percutaneous tumor biopsy was performed using CT fluoroscopy with reference to the MR images obtained beforehand. The sampling of multiple biopsy specimens was successful by the coaxial method using a 16/18G Quick-Core biopsy needle in all the cases, and definitive histopathologic diagnoses were achieved. The final pathological diagnoses were one benign lesion (pseudowidening of adenomyosis) and two malignant lesions (small cell carcinoma and adenocarcinoma of Mullerian origin of ovaries). There were no associated complications. The treatment strategy was formulated according to the results of the pathological diagnosis in each case, and consisted of either hormonal therapy or systemic chemotherapy.

The confusion of MRI sequence names could be solved if MR images were automatically identified after image data acquisition. We revealed the ability of deep learning to classify head MRI sequences. Seventy-eight patients with mild cognitive impairment (MCI) having apparently normal head MR images and 78 intracranial hemorrhage (ICH) patients with morphologically deformed head MR images were enrolled. Six imaging protocols were selected to be performed: T2-weighted imaging, fluid attenuated inversion recovery imaging, T2-star-weighted imaging, diffusion-weighted imaging, apparent diffusion coefficient mapping, and source images of time-of-flight magnetic resonance angiography. The proximal first image slices and middle image slices having ambiguous and distinctive contrast patterns, respectively, were classified by two deep learning imaging classifiers, AlexNet and GoogLeNet. AlexNet had accuracies of 73.3%, 73.6%, 73.1%, and 60.7% in the middle slices of MCI group, middle slices of ICH group, first slices of MCI group, and first slices of ICH group, while GoogLeNet had accuracies of 100%, 98.1%, 93.1%, and 94.8%, respectively. AlexNet significantly had lower classification ability than GoogLeNet for all datasets. GoogLeNet could judge the types of head MRI sequences with a small amount of training data, irrespective of morphological or contrast conditions.

In-111 ibritumomab tiuxetan pretherapy imaging is performed prior to radioimmunotherapy with Y-90 ibritumomab tiuxetan in patients with lymphoma in order to estimate the biodistribution of the radiolabeled antibodies. We report the case of a 61-year-old woman with relapsed follicular lymphoma. In-111 ibritumomab tiuxetan pretherapy imaging showed tracer accumulation in a surgically treated fracture of the left calcaneus as well as in the involved lesions. The patient had fractured her left calcaneus 6 months before the radioimmunotherapy, and fracture of the left calcaneus without lymphoma involvement had been revealed operatively at that time. Clinical features and magnetic resonance imaging (MRI) indicated an inflammatory response in the left heel. We concluded that this inflammation played an important role in the uptake on pretherapy imaging. It should be kept in mind that the inflammation is a differential diagnosis of lymphoma involvement in In-111 ibritumomab tiuxetan pretherapy imaging. To the best of our knowledge, this is the first case report of In-111 ibritumomab tiuxetan uptake in a surgically treated fracture.

Background: Recent studies have reported the clinical usefulness of the soft-copy reading of mammograms. However, image-processing parameters for soft-copy reading of digital mammograms have not been established. Purpose: To compare observer performance in detecting breast cancer by soft-copy reading of digital mammograms using a routine image-processing parameter versus each of several high-contrast parameters. Material and Methods: The mammograms of 154 breasts, including 48 abnormal breasts with breast cancer and 106 normal breasts, were examined. Cancers were classified into 34 mass-dominant cancers, 11 microcalcification-dominant cancers, two cancers showing only architectural distortion, and one cancer without abnormal findings. All mammograms were performed using a computed radiography (CR) system. Each image was processed using GA (1.2), which was the contrast parameter recommended by the manufacturer for hard-copy film, GA (1.4), GA (1.6), and GA (1.8). These images were displayed on 5-megapixel (M) liquid-crystal display monitors. Five experienced radiologists classified them into BI-RADS category 1–2 or 3–5, and were also asked to rate the images on a scale of 0 to 100 for the likelihood of the presence of masses and microcalcifications in each breast. Results: In mass-dominant cancers of dense breast tissue, the mean sensitivities of GA (1.2), GA (1.4), GA (1.6), and GA (1.8) were 32.7, 38.2, 36.4, and 40.0, and the AZ values were 0.67, 0.73, 0.71, and 0.73, respectively; in microcalcification-dominant cancers, the mean sensitivities were 80.0, 74.5, 80.0, and 78.2, respectively; however, there were no significant differences among them. Conclusion: High-contrast parameters tended to show relatively high sensitivity and AZ values in the detection of masses in dense breast tissue, but relatively low sensitivity for microcalcifications.

We report two cases of inflammatory myofibroblastic tumor (IMT) of the lung in a 4-year-old boy and a 7-year-old girl. We performed dynamic contrastenhancement computed tomography in both of our cases and dynamic contrast enhancement magnetic resonance imaging (MRI) in one case. These dynamic studies showed a demarcated mass with delayed enhancement in both cases. A T1-weighted MR image shows a mass with homogeneous low signal intensity, and a T2- weighted image shows a mass with slightly high signal intensity. A post-contrast-enhanced T1-weighted image demonstrates homogeneous enhancement. We speculate that the delayed enhancement could be attributed to the abundant fibrous tissue, which was the main structural material of the tumor. This same finding has been described in previous reports of IMT in other organs.

To determine the value of adding diffusion-weighted (DW) magnetic resonance (MR) imaging to dynamic contrast material-enhanced MR imaging when distinguishing between benign and malignant parotid tumors.This retrospective study was approved by the institutional review board, and the informed consent requirement was waived. The authors analyzed MR images of 50 lesions (36 benign, 14 malignant) in 47 patients. DW MR imaging and dynamic contrast-enhanced MR imaging were performed in all patients. Time-intensity curve (TIC) patterns were categorized as follows: type A, time to peak was more than 120 seconds; type B, time to peak was 120 seconds or less with high washout ratio (or = 30%); type C, time to peak was 120 seconds or less with low washout ratio (30%); and type D, flat. The apparent diffusion coefficient (ADC) values were measured on DW MR images. Sensitivity, specificity, accuracy, and positive and negative predictive values were calculated for type A, B, and D tumors regarded as benign and for type C tumors regarded as malignant. On the basis of DW MR imaging results, ADC threshold values between pleomorphic adenomas and carcinomas and between Warthin tumors and carcinomas were selected. Diagnostic accuracy was compared before and after modification diagnosis referring to the ADC value obtained with the McNemar test. P.05 was considered to indicate a significant difference.ADC threshold values were 1.4 x 10(-3) mm(2)/sec between pleomorphic adenomas and carcinomas and 1.0 x 10(-3) mm(2)/sec between Warthin tumors and carcinomas. Accuracy (82% vs 94%) and positive predictive value (67% vs 92%) significantly improved with the addition of ADC values in the evaluation of patients with type B or C tumors.A persistent or flat TIC pattern on dynamic contrast-enhanced MR images indicates benign disease, but there is added value from including the ADC value in the evaluation of tumors that show a plateau or washout TIC pattern.

Purpose: We evaluated the magnetic resonance (MR) features of breast lesions showing circumscribed mass on mammography to understand the characteristics that differentiate malignancy and benignancy. Materials and Methods: Our institutional review board approved the study, and informed consent was waived. Using logistic regression analysis, we examined morphologic and kinetic MR imaging data of 90 breast lesions (43 malignant, 47 benign) that showed circumscribed mass on mammography. Results: Features identified as having high odds for malignancy included: rim enhancement (odds ratio, 70.894; 95% confidence interval (CI), 7.525-667.938); heterogeneous enhancement (odds ratio, 10.839; 95% CI, 1.032-113.856); and washout dynamic pattern (odds ratio, 46.262; 95% CI, 3.716-575.901). Combinations of washout dynamic pattern and either rim or heterogeneous enhancement reflected excessively high prediction probability for malignancy (>0.95), whereas combinations lacking washout dynamic pattern and with either homogeneous enhancement or dark internal septation revealed excessively low prediction probability for malignancy (

The case of a 46-year-old woman with lymphangioleiomyomatosis (LAM) involving the supraclavicular, mediastinal, and pelvic lymph nodes in addition to the lungs is reported. Computed tomography incidentally revealed multiple thin-walled pulmonary cysts and low-attenuating masses in the supraclavicular, mediastinal, and retroperitoneal lymph nodes. A biopsy of the supraclavicular mass was performed and diagnosed as LAM histopathologically. The common sites of extrapulmonary LAM include retroperitoneal and mediastinal lymph nodes; however, supraclavicular lymph node involvement is extremely rare.

We report a case of Gorham's disease of the chest wall in a 29-year-old man. A chest radiograph showed osteolysis of the left first, second, and third ribs. CT and MR images revealed a soft tissue mass around the affected ribs. The mass presented as having heterogeneous high signal intensity on T2-weighted MR images, and it showed minimal enhancement on contrast-enhanced MR images. A histologic specimen obtained by incisional biopsy from the left third rib revealed proliferation of lymphangiomatous tissue in the bone marrow. A diagnosis of Gorham's disease was made based on these clinical, radiologic, and histologic findings.

We compared observer performance of digital mammography among hard-copy readings and soft-copy readings using 3-megapixel (3M) and 5-megapixel (5M) liquid crystal display (LCD) monitors. Five experienced radiologists assessed 80 mammograms of 40 cancers and 40 benign lesions. There were no significant differences among the average A(z) of three modalities and among the κ values for intra- and interobserver agreement. The soft-copy reading using the 3M LCD monitor took a slightly longer time, although there were no significant differences.

Purpose To evaluate the diagnostic accuracy of a combination of dynamic contrast-enhanced MR imaging (DCE-MRI) and diffusion-weighted MR imaging (DWI) in characterization of lesions showing non-mass-like enhancement on breast MR imaging and to find the strongest discriminators between carcinoma and benignancy. Materials and methods We analyzed consecutive MR images in 45 lesions showing non-mass like enhancement in 41 patients. We analyzed lesion size, distribution, internal enhancement, kinetic curve pattern, and apparent diffusion coefficient (ADC) values. We applied univariate and multivariate analyses to find the strongest indicators for malignancy. In a validation study, 22 non-mass-like enhancement lesions in 21 patients were examined. We calculated diagnostic accuracy when we presume category 4b, 4c, and 5 lesions as malignant or high to moderate suspicion for malignancy, and category 4a and 3 as low suspicion for malignancy or benign. Results Segmental distribution (P = 0.018), clumped internal enhancement (P = 0.005), and ADC less than 1.3 × 10−3 mm2/s (P = 0.047) were the strongest MR indicators of malignancy. In a validation study, sensitivity, specificity, positive predictive value, negative predictive value and accuracy were 87% (13/15), 86% (6/7), 93% (13/14), 75% (6/8) and 86% (19/22), respectively. Conclusion The combination of DCE-MRI and DWI showed high diagnostic accuracy in characterization of non-mass-like enhancement lesions on breast MR images.

Purpose To investigate optimal gradation processing parameter for soft-copy reading comparing the parameter recommended for hard-copy with other parameters. Materials and methods Digital mammograms using a Computed Radiography system were evaluated. The gradation processing parameter recommended by the manufacturer was GA (1.2). We prepared seven parameters that changed the degree of contrast: GA (0.6, 0.8, 1.0, 1.2, 1.4, 1.6, and 1.8). Images of an anthropomorphic breast phantom were displayed on a 5-megapixel liquid crystal display monitor using each parameter. Three readers independently assessed each image, and scored for the following items: intramammary contrast, extramammary contrast, sharpness, and graininess. Total score was calculated in each parameter. We also displayed normal mammograms of nine cases: three with a scattered fibroglandular density, three with a heterogeneously dense breast, and three with an extremely dense breast. These were displayed using GA (1.2) or using parameters with a higher total score than GA (1.2) in the phantom test. Three readers assessed each mammogram as in the phantom test. Results In phantom test, GA (1.4), GA (1.6), and GA (1.8) obtained higher scores than GA (1.2). In normal cases tested using these parameters, GA (1.4) obtained the highest score. This was significantly higher than that of GA (1.2) ( P = 0.004). The score obtained for GA (1.4) was the highest in cases with extremely dense and heterogeneously dense breast tissue, though there was no statistically significant difference. Conclusion Soft-copy image quality was improved by gradient processing using higher contrast parameter than that routinely used in hard-copy, especially in dense breast cases.

To determine whether muscle fiber atrophy associated with steroid myopathy can be detected with T2 relaxation time.Animal and human studies were approved by the ethics committee. Informed consent was obtained. Twelve rabbits were divided into a group that received 3 mg/kg of triamcinolone subcutaneously each day for 10 consecutive days (n = 6) and a control group that received saline (n = 6). Magnetic resonance (MR) imaging was performed before and after treatment. T2 and fat deposition ratio (FDR) of soleus and gastrocnemius muscles before and after treatment and between control rabbits and rabbits treated with steroids were compared by using two-way repeated analysis of variance and Bonferroni post hoc test to evaluate effects of steroid treatment. After imaging, rabbits were sacrificed. Extracellular space ratio (ECSR) and fiber diameter were examined. Correlation among T2, ECSR, and diameter of type 2 muscle fibers was analyzed with a Pearson correlation test with Bonferroni correction in gastrocnemius to determine factors affecting T2. In humans, T2 relaxation time and FDR of both muscles were compared between volunteers not treated with steroids and patients treated with steroids by using an unpaired t test to evaluate the effects of steroids.In rabbits, T2 of gastrocnemius muscle was significantly (P.01) longer after steroid treatment than before steroid treatment and was also significantly (P.01) longer than after saline administration. T2 of the gastrocnemius showed no significant difference in control rabbits before or after saline administration or in control rabbits and rabbits before steroid administration. T2 of the soleus muscle or FDR of either muscle showed no significant difference. There was a significant correlation (P.01) among T2, ECSR, and diameter of type 2 muscle fibers in the gastrocnemius. In humans, T2 of the gastrocnemius was significantly (P.01) longer in patients than in volunteers. T2 of the soleus or FDR of either muscle showed no significant difference.Muscle fiber atrophy associated with steroid myopathy is detectable as prolongation of T2 relaxation time in the gastrocnemius muscle; the authors believe prolongation of T2 relaxation time is mainly due to increased ECSR reflecting type 2 muscle fiber atrophy.

CT and MR findings of a case of brain aspergillosis with histopathologic correlation are reported. On both CT and MR images, there were multiple lesions in the corticomedullary junction (CMJ) that appeared to disrupt the cortical sulci and that were not enhanced by intravenous contrast material. In most of these lesions, there were centrally located structures that were enhanced by intravenous contrast material and that appeared to be continuous from markedly enhanced adjacent dilated cortical vessels. Histopathologic examination of the autopsy specimen showed multiple hemorrhagic infarcts in the CMJ with remaining dilated cortical vessels that had been thrombosed by aspergillus hyphae.