Deep-learning model for diagnostic clue: detecting the dural tail sign for meningiomas on contrast-enhanced T1 weighted images.

Background: Meningiomas are the most common primary central nervous system tumors, and magnetic resonance imaging (MRI), especially contrast-enhanced T1 weighted image (CE T1WI), is used as a fundamental imaging modality for the detection and analysis of the tumors. In this study, we propose an automated deep-learning model for meningioma detection using the dural tail sign. Methods: The dataset included 123 patients with 3,824 dural tail signs on sagittal CE T1WI. The dataset was divided into training and test datasets based on specific time point, and 78 and 45 patients were comprised for the training and test dataset, respectively. To compensate for the small sample size of the training dataset, 39 additional patients with 69 dural tail signs from the open dataset were appended to the training dataset. A You Only Look Once (YOLO) v4 network was trained with sagittal CE T1WI to detect dural tail signs. The normal group dataset, comprised of 51 patients with no abnormal finding on MRI, was employed to evaluate the specificity of the trained model. Results: The sensitivity and false positive average were 82.22% and 29.73, respectively, in the test dataset. The specificity and false positive average were 17.65% and 3.16, respectively, in the normal dataset. Most of the false-positive cases in the test dataset were enhancing vessels, misinterpreted as dural thickening. Conclusions: The proposed model demonstrates an automated detection system for the dural tail sign to identify meningioma in general screening MRI. Our model can facilitate and alleviate radiologists' reading process by notifying the possibility of incidental dural mass based on dural tail sign detection.

Developing and Evaluating Deep Learning Algorithms for Object Detection: Key Points for Achieving Superior Model Performance.

In recent years, artificial intelligence, especially object detection-based deep learning in computer vision, has made significant advancements, driven by the development of computing power and the widespread use of graphic processor units. Object detection-based deep learning techniques have been applied in various fields, including the medical imaging domain, where remarkable achievements have been reported in disease detection. However, the application of deep learning does not always guarantee satisfactory performance, and researchers have been employing trial-and-error to identify the factors contributing to performance degradation and enhance their models. Moreover, due to the black-box problem, the intermediate processes of a deep learning network cannot be comprehended by humans; as a result, identifying problems in a deep learning model that exhibits poor performance can be challenging. This article highlights potential issues that may cause performance degradation at each deep learning step in the medical imaging domain and discusses factors that must be considered to improve the performance of deep learning models. Researchers who wish to begin deep learning research can reduce the required amount of trial-and-error by understanding the issues discussed in this study.

Gray-white matter boundary Z-score and volume as imaging biomarkers of Alzheimer's disease.

Introduction: Alzheimer's disease (AD) presents typically gray matter atrophy and white matter abnormalities in neuroimaging, suggesting that the gray-white matter boundary could be altered in individuals with AD. The purpose of this study was to explore differences of gray-white matter boundary Z-score (gwBZ) and its tissue volume (gwBTV) between patients with AD, amnestic mild cognitive impairment (MCI), and cognitively normal (CN) elderly participants. Methods: Three-dimensional T1-weight images of a total of 227 participants were prospectively obtained from our institute from 2006 to 2022 to map gwBZ and gwBTV on images. Statistical analyses of gwBZ and gwBTV were performed to compare the three groups (AD, MCI, CN), to assess their correlations with age and Korean version of the Mini-Mental State Examination (K-MMSE), and to evaluate their effects on AD classification in the hippocampus. Results: This study included 62 CN participants (71.8 ± 4.8 years, 20 males, 42 females), 72 MCI participants (72.6 ± 5.1 years, 23 males, 49 females), and 93 AD participants (73.6 ± 7.7 years, 22 males, 71 females). The AD group had lower gwBZ and gwBTV than CN and MCI groups. K-MMSE showed positive correlations with gwBZ and gwBTV whereas age showed negative correlations with gwBZ and gwBTV. The combination of gwBZ or gwBTV with K-MMSE had a high accuracy in classifying AD from CN in the hippocampus with an area under curve (AUC) value of 0.972 for both. Conclusion: gwBZ and gwBTV were reduced in AD. They were correlated with cognitive function and age. Moreover, gwBZ or gwBTV combined with K-MMSE had a high accuracy in differentiating AD from CN in the hippocampus. These findings suggest that evaluating gwBZ and gwBTV in AD brain could be a useful tool for monitoring AD progression and diagnosis.

Deep learning-based detection algorithm for brain metastases on black blood imaging.

Brain metastases (BM) are the most common intracranial tumors, and their prevalence is increasing. High-resolution black-blood (BB) imaging was used to complement the conventional contrast-enhanced 3D gradient-echo imaging to detect BM. In this study, we propose an efficient deep learning algorithm (DLA) for BM detection in BB imaging with contrast enhancement scans, and assess the efficacy of an automatic detection algorithm for BM. A total of 113 BM participants with 585 metastases were included in the training cohort for five-fold cross-validation. The You Only Look Once (YOLO) V2 network was trained with 3D BB sampling perfection with application-optimized contrasts using different flip angle evolution (SPACE) images to investigate the BM detection. For the observer performance, two board-certified radiologists and two second-year radiology residents detected the BM and recorded the reading time. For the training cohort, the overall performance of the five-fold cross-validation was 87.95%, 24.82%, 19.35%, 14.48, and 18.40 for sensitivity, precision, F1-Score, the false positive average for the BM dataset, and the false positive average for the normal individual dataset, respectively. For the comparison of reading time with and without DLA, the average reading time was reduced by 20.86% in the range of 15.22-25.77%. The proposed method has the potential to detect BM with a high sensitivity and has a limited number of false positives using BB imaging.

Reliable quality assurance of X-ray mammography scanner by evaluation the standard mammography phantom image using an interpretable deep learning model.

OBJECTIVE: Mammography is the initial examination to detect breast cancer symptoms, and quality control of mammography devices is crucial to maintain accurate diagnosis and to safeguard against degradation of performance. The objective of this study was to assist radiologists in mammography phantom image evaluation by developing and validating an interpretable deep learning model capable of objectively evaluating the quality of standard phantom images for mammography. MATERIALS AND METHODS: A total of 2,208 mammography phantom images were collected for periodic accreditation of the scanner from 1,755 institutions. The dataset was randomly split into training (1,808 images) and testing (400 images) datasets with subgroups (76 images) for the multi-reader study. To develop an interpretable model that contains two deep learning networks in series, five processing steps were performed: mammography phantom detection, phantom object detection, post-processing, score evaluation, and a report with a comment about ambiguous results. RESULTS: For phantom detection, the accuracy and mean intersection over union (mIOU) were 1.00 and 0.938 in the test dataset, respectively. During phantom object detection, a total of 6,369 out of 6,400 objects were detected as the correct object class, and the accuracy and mIOU were 0.995 and 0.813, respectively. The predicted score for each object showed a consensus of 97.40% excluding ambiguous points and 59.10% for ambiguous points of the groups. CONCLUSIONS: The interpretable deep learning model using large-scale data from multiple centers shows high performance and reasonable object scoring, successfully validating the reliability and feasibility of mammography phantom image quality management.

Sex-Specific Differences in Lower Body Fat Distribution and Association with Physical Performance among Healthy Community-Dwelling Older Adults: A Pilot Study.

This study aims to examine sex-specific differences in body composition and lower extremity fat distribution and their association with physical performance among healthy older adults. The pilot study comprises 40 subjects (20 men and 20 women) matched by age and body mass index. The participants undergo dual-energy X-ray absorptiometry, magnetic resonance imaging, and proton magnetic resonance spectroscopy (1H-MRS) to assess body composition and lower extremity fat distribution. 1H-MRS is used to measure the extramyocellular lipid (EMCL) and intramyocellular lipid (IMCL) contents of the lower leg muscles (soleus and tibialis anterior) at the maximum circumference of the calf after overnight fasting. The tibialis anterior IMCL, as assessed by 1H-MRS, is negatively associated with the five-times sit-to-stand test scores (rs = 0.518, p = 0.023) in men, while the soleus IMCL content is negatively associated with the timed up-and-go test scores (rs = 0.472, p = 0.048) in women. However, the soleus EMCL content is positively associated with the five-times sit-to-stand test scores (rs = -0.488, p = 0.040) in women, but this association is not statistically significant in men. This study shows an inverse correlation between IMCL content and physical performance in healthy older individuals and lower leg muscle-specific IMCL based on sex differences. Furthermore, our results suggest that greater EMCL content in the soleus and calf subcutaneous fat might affect physical performance positively in women but not men.

Effective End-to-End Deep Learning Process in Medical Imaging Using Independent Task Learning: Application for Diagnosis of Maxillary Sinusitis.

PURPOSE: This study aimed to propose an effective end-to-end process in medical imaging using an independent task learning (ITL) algorithm and to evaluate its performance in maxillary sinusitis applications. MATERIALS AND METHODS: For the internal dataset, 2122 Waters' view X-ray images, which included 1376 normal and 746 sinusitis images, were divided into training (n=1824) and test (n=298) datasets. For external validation, 700 images, including 379 normal and 321 sinusitis images, from three different institutions were evaluated. To develop the automatic diagnosis system algorithm, four processing steps were performed: 1) preprocessing for ITL, 2) facial patch detection, 3) maxillary sinusitis detection, and 4) a localization report with the sinusitis detector. RESULTS: The accuracy of facial patch detection, which was the first step in the end-to-end algorithm, was 100%, 100%, 99.5%, and 97.5% for the internal set and external validation sets #1, #2, and #3, respectively. The accuracy and area under the receiver operating characteristic curve (AUC) of maxillary sinusitis detection were 88.93% (0.89), 91.67% (0.90), 90.45% (0.86), and 85.13% (0.85) for the internal set and external validation sets #1, #2, and #3, respectively. The accuracy and AUC of the fully automatic sinusitis diagnosis system, including site localization, were 79.87% (0.80), 84.67% (0.82), 83.92% (0.82), and 73.85% (0.74) for the internal set and external validation sets #1, #2, and #3, respectively. CONCLUSION: ITL application for maxillary sinusitis showed reasonable performance in internal and external validation tests, compared with applications used in previous studies.

Added Value of Chemical Exchange-Dependent Saturation Transfer MRI for the Diagnosis of Dementia.

OBJECTIVE: Chemical exchange-dependent saturation transfer (CEST) MRI is sensitive for detecting solid-like proteins and may detect changes in the levels of mobile proteins and peptides in tissues. The objective of this study was to evaluate the characteristics of chemical exchange proton pools using the CEST MRI technique in patients with dementia. MATERIALS AND METHODS: Our institutional review board approved this cross-sectional prospective study and informed consent was obtained from all participants. This study included 41 subjects (19 with dementia and 22 without dementia). Complete CEST data of the brain were obtained using a three-dimensional gradient and spin-echo sequence to map CEST indices, such as amide, amine, hydroxyl, and magnetization transfer ratio asymmetry (MTRasym) values, using six-pool Lorentzian fitting. Statistical analyses of CEST indices were performed to evaluate group comparisons, their correlations with gray matter volume (GMV) and Mini-Mental State Examination (MMSE) scores, and receiver operating characteristic (ROC) curves. RESULTS: Amine signals (0.029 for non-dementia, 0.046 for dementia, p = 0.011 at hippocampus) and MTRasym values at 3 ppm (0.748 for non-dementia, 1.138 for dementia, p = 0.022 at hippocampus), and 3.5 ppm (0.463 for non-dementia, 0.875 for dementia, p = 0.029 at hippocampus) were significantly higher in the dementia group than in the non-dementia group. Most CEST indices were not significantly correlated with GMV; however, except amide, most indices were significantly correlated with the MMSE scores. The classification power of most CEST indices was lower than that of GMV but adding one of the CEST indices in GMV improved the classification between the subject groups. The largest improvement was seen in the MTRasym values at 2 ppm in the anterior cingulate (area under the ROC curve = 0.981), with a sensitivity of 100 and a specificity of 90.91. CONCLUSION: CEST MRI potentially allows noninvasive image alterations in the Alzheimer's disease brain without injecting isotopes for monitoring different disease states and may provide a new imaging biomarker in the future.

Lifestyle-dependent brain change: a longitudinal cohort MRI study.

We investigated both independent and interconnected effects of 3 lifestyle factors on brain volume, measuring yearly changes using large-scale longitudinal magnetic resonance imaging, in middle-aged to older adults. We measured brain volumes in a cohort (n = 984, 49-79 years) from the Korean Genome and Epidemiology Study group, using baseline and follow-up estimates after 4 years. In our analysis, the accelerated brain atrophy in normal aging was observed across regions (e.g., brain tissue: -0.098 ± 0.01 mL/y, p < 0.001). An independent lifestyle-specific trend of brain atrophy across time was also evident in men, where smoking (p = 0.012) and physical activity (p = 0.014) showed the strongest association with the atrophy rate. Linear regression analysis of the interconnected effect revealed that brain atrophy is mitigated by intense physical activity in smoking males. Lifestyle factors did not show any significant effect on brain volume in women. These results provide important information regarding lifestyle factors that affect brain aging in mid-to-late adulthood. Our findings may aid in the identification of preventive measures against dementia.

Estimation of T2* Relaxation Time of Breast Cancer: Correlation with Clinical, Imaging and Pathological Features.

OBJECTIVE: The purpose of this study was to estimate the T2* relaxation time in breast cancer, and to evaluate the association between the T2* value with clinical-imaging-pathological features of breast cancer. MATERIALS AND METHODS: Between January 2011 and July 2013, 107 consecutive women with 107 breast cancers underwent multi-echo T2*-weighted imaging on a 3T clinical magnetic resonance imaging system. The Student's t test and one-way analysis of variance were used to compare the T2* values of cancer for different groups, based on the clinical-imaging-pathological features. In addition, multiple linear regression analysis was performed to find independent predictive factors associated with the T2* values. RESULTS: Of the 107 breast cancers, 92 were invasive and 15 were ductal carcinoma in situ (DCIS). The mean T2* value of invasive cancers was significantly longer than that of DCIS (p = 0.029). Signal intensity on T2-weighted imaging (T2WI) and histologic grade of invasive breast cancers showed significant correlation with T2* relaxation time in univariate and multivariate analysis. Breast cancer groups with higher signal intensity on T2WI showed longer T2* relaxation time (p = 0.005). Cancer groups with higher histologic grade showed longer T2* relaxation time (p = 0.017). CONCLUSION: The T2* value is significantly longer in invasive cancer than in DCIS. In invasive cancers, T2* relaxation time is significantly longer in higher histologic grades and high signal intensity on T2WI. Based on these preliminary data, quantitative T2* mapping has the potential to be useful in the characterization of breast cancer.

Prostate cancer: correlation of intravoxel incoherent motion MR parameters with Gleason score.

PURPOSE: To evaluate the potential of intravoxel incoherent motion (IVIM) imaging to predict histological prognostic parameters by investigating whether IVIM parameters correlate with Gleason score. MATERIALS AND METHODS: The institutional review board approved this retrospective study, and informed consent was waived. A total of 41 patients with histologically proven prostate cancer who underwent prostate MRI using a 3T MRI machine were included. For eight diffusion-weighted imaging b-values (0, 10, 20, 50, 100, 200, 500, and 800s/mm(2)), a spin-echo echo-planar imaging sequence was performed. D, f, D(⁎), and ADCfit values were compared among three groups of patients with prostate cancer: Gleason score 6 (n=9), 7 (n=16), or 8 or higher (n=16). Receiver operating characteristic (ROC) curves were generated for D, f, D(⁎), and ADCfit to assess the ability of each parameter to distinguish cancers with low Gleason scores from cancers with intermediate or high Gleason scores. RESULTS: Pearson's coefficient analysis revealed significant negative correlations between Gleason score and ADCfit (r=-0.490, P=.001) and Gleason score and D values (r=-0.514, P=.001). Gleason score was poorly correlated with f (r=0.168, P=.292) and D(⁎) values (r=-0.108, P=.500). The ADCfit and D values of prostate cancers with Gleason scores 7 or ≥8 were significantly lower than values for prostate cancers with Gleason score 6 (P<.05). ROC curves were constructed to assess the ability of IVIM parameters to discriminate prostate cancers with Gleason score 6 from cancers with Gleason scores 7 or ≥8. Areas under the curve were 0.671 to 0.974. ADCfit and D yielded the highest Az value (0.960-0.956), whereas f yielded the lowest Az value (0.633). CONCLUSIONS: The pure molecular diffusion parameter, D, was the IVIM parameter that best discriminated prostate cancers with low Gleason scores from prostate cancers with intermediate or high Gleason scores.