Coronal plane in opportunistic screening of osteoporosis using computed tomography: comparison with axial and sagittal planes.

OBJECTIVE: To compare the coronal plane with axial and sagittal planes in opportunistic screening of osteoporosis using computed tomography (CT). MATERIALS AND METHODS: A total of 100 patients aged ≥ 50 years who underwent both lumbar spine CT and dual-energy X-ray absorptiometry within 3 months were included. Osteoporosis was diagnosed based on dual-energy X-ray absorptiometry results. The CT number was measured at the center of the vertebral body in coronal, axial, and sagittal planes. To compare the coronal plane with axial and sagittal planes in diagnosing osteoporosis, the areas under the receiver operating characteristic curve (AUC) were compared and intraclass correlation coefficient (ICC) was calculated. The optimal cutoff values were calculated using Youden's index. RESULTS: The AUC of the coronal plane (0.80; 95% confidence interval [CI], 0.71-0.89) was not significantly different from that of the axial plane (0.78; 95% CI, 0.68-0.87; P = 0.39) and that of the sagittal plane (0.78; 95% CI, 0.69-0.87; P = 0.68). Excellent concordance rates were observed between coronal and axial planes with ICC of 0.95 (95% CI, 0.92-0.96) and between coronal and sagittal planes with ICC of 0.93 (95% CI, 0.85-0.96). The optimal cutoff values for the coronal, axial, and sagittal planes were 110, 112, and 112 HU, respectively. CONCLUSION: The coronal plane does not significantly differ from axial and sagittal planes in opportunistic screening of osteoporosis. Thus, the coronal plane as well as axial and sagittal planes can be used interchangeably in measuring bone mineral density using CT.

Intermediate valence state in YbB4revealed by resonant x-ray emission spectroscopy.

We report the temperature dependence of the Yb valence in the geometrically frustrated compoundYbB4from 12 to 300 K using resonant x-ray emission spectroscopy at the YbLα1transition. We find that the Yb valence,v, is hybridized between thev = 2 andv = 3 valence states, increasing fromv=2.61±0.01at 12 K tov=2.67±0.01at 300 K, confirming thatYbB4is a Kondo system in the intermediate valence regime. This result indicates that the Kondo interaction inYbB4is substantial, and is likely to be the reason whyYbB4does not order magnetically at low temperature, rather than this being an effect of geometric frustration. Furthermore, the zero-point valence of the system is extracted from our data and compared with other Kondo lattice systems. The zero-point valence seems to be weakly dependent on the Kondo temperature scale, but not on the valence change temperature scaleTv.

Appendiceal Visualization on 2-mSv CT vs. Conventional-Dose CT in Adolescents and Young Adults with Suspected Appendicitis: An Analysis of Large Pragmatic Randomized Trial Data.

OBJECTIVE: We compared appendiceal visualization on 2-mSv CT vs. conventional-dose CT (median 7 mSv) in adolescents and young adults and analyzed the undesirable clinical and diagnostic outcomes that followed appendiceal nonvisualization. MATERIALS AND METHODS: A total of 3074 patients aged 15-44 years (mean ± standard deviation, 28 ± 9 years; 1672 female) from 20 hospitals were randomized to the 2-mSv CT or conventional-dose CT group (1535 vs. 1539) from December 2013 through August 2016. A total of 161 radiologists from 20 institutions prospectively rated appendiceal visualization (grade 0, not identified; grade 1, unsure or partly visualized; and grade 2, clearly and entirely visualized) and the presence of appendicitis in these patients. The final diagnosis was based on CT imaging and surgical, pathologic, and clinical findings. We analyzed undesirable clinical or diagnostic outcomes, such as negative appendectomy, perforated appendicitis, more extensive than simple appendectomy, delay in patient management, or incorrect CT diagnosis, which followed appendiceal nonvisualization (defined as grade 0 or 1) and compared the outcomes between the two groups. RESULTS: In the 2-mSv CT and conventional-dose CT groups, appendiceal visualization was rated as grade 0 in 41 (2.7%) and 18 (1.2%) patients, respectively; grade 1 in 181 (11.8%) and 81 (5.3%) patients, respectively; and grade 2 in 1304 (85.0%) and 1421 (92.3%) patients, respectively (p < 0.001). Overall, undesirable outcomes were rare in both groups. Compared to the conventional-dose CT group, the 2-mSv CT group had slightly higher rates of perforated appendicitis (1.1% [17] vs. 0.5% [7], p = 0.06) and false-negative diagnoses (0.4% [6] vs. 0.0% [0], p = 0.01) following appendiceal nonvisualization. Otherwise, these two groups were comparable. CONCLUSION: The use of 2-mSv CT instead of conventional-dose CT impairs appendiceal visualization in more patients. However, appendiceal nonvisualization on 2-mSv CT rarely leads to undesirable clinical or diagnostic outcomes.

Magnetic dilution of a honeycomb lattice XY magnet CoTiO3.

We report our study of cobalt (II) titanate, CoTiO3, in which magnetic Co ions are replaced by non-magnetic ions. The antiferromagnetic ordering transition of CoTiO3around 37 K is described with ferromagnetic honeycomb layers coupled antiferromagnetically along the crystallographicc-direction. The effect of magnetic dilution on the Néel temperature of this material is investigated through the doping of Zn2+and Mg2+in place of Co2+for various dilution levels up tox+y= 0.46 in Co1-x-yZnxMgyTiO3. Single phase polycrystalline samples have been synthesized and their structural and magnetic properties have been examined. A linear relation between dilution and the Néel temperature is observed over a wide doping range. A linear extrapolation would suggest that the required dilution level to suppress magnetic order is aroundx+y∼ 0.74, well beyond the classical percolation threshold. The implication of this observation for microscopic models for describing CoTiO3is discussed.

Diagnostic accuracy and complication rate of image-guided percutaneous transthoracic needle lung biopsy for subsolid pulmonary nodules: a systematic review and meta-analysis.

OBJECTIVES: To determine the diagnostic accuracy and complication rate of percutaneous transthoracic needle biopsy (PTNB) for subsolid pulmonary nodules and sources of heterogeneity among reported results. METHODS: We searched PubMed, EMBASE, and Cochrane libraries (until November 7, 2020) for studies measuring the diagnostic accuracy of PTNB for subsolid pulmonary nodules. Pooled sensitivity and specificity of PTNB were calculated using a bivariate random-effects model. Bivariate meta-regression analyses were performed to identify sources of heterogeneity. Pooled overall and major complication rates were calculated. RESULTS: We included 744 biopsies from 685 patients (12 studies). The pooled sensitivity and specificity of PTNB for subsolid nodules were 90% (95% confidence interval [CI]: 85-94%) and 99% (95% CI: 92-100%), respectively. Mean age above 65 years was the only covariate significantly associated with higher sensitivity (93% vs 85%, p = 0.04). Core needle biopsy showed marginally higher sensitivity than fine-needle aspiration (93% vs 83%, p = 0.07). Pooled overall and major complication rate of PTNB were 43% (95% CI: 25-62%) and 0.1% (95% CI: 0-0.4%), respectively. Major complication rate was not different between fine-needle aspiration and core needle biopsy groups (p = 0.25). CONCLUSION: PTNB had acceptable performance and a low major complication rate in diagnosing subsolid pulmonary nodules. The only significant source of heterogeneity in reported sensitivities was a mean age above 65 years. ADVANCES IN KNOWLEDGE: This is the first meta-analysis attempting to systemically determine the cause of heterogeneity in the diagnostic accuracy and complication rate of PTNB for subsolid pulmonary nodules.

Deep learning-based denoising algorithm in comparison to iterative reconstruction and filtered back projection: a 12-reader phantom study.

OBJECTIVES: (1) To compare low-contrast detectability of a deep learning-based denoising algorithm (DLA) with ADMIRE and FBP, and (2) to compare image quality parameters of DLA with those of reconstruction methods from two different CT vendors (ADMIRE, IMR, and FBP). MATERIALS AND METHODS: Using abdominal CT images of 100 patients reconstructed via ADMIRE and FBP, we trained DLA by feeding FBP images as input and ADMIRE images as the ground truth. To measure the low-contrast detectability, the randomized repeat scans of Catphan® phantom were performed under various conditions of radiation exposures. Twelve radiologists evaluated the presence/absence of a target on a five-point confidence scale. The multi-reader multi-case area under the receiver operating characteristic curve (AUC) was calculated, and non-inferiority tests were performed. Using American College of Radiology CT accreditation phantom, contrast-to-noise ratio, target transfer function, noise magnitude, and detectability index (d') of DLA, ADMIRE, IMR, and FBPs were computed. RESULTS: The AUC of DLA in low-contrast detectability was non-inferior to that of ADMIRE (p < .001) and superior to that of FBP (p < .001). DLA improved the image quality in terms of all physical measurements compared to FBPs from both CT vendors and showed profiles of physical measurements similar to those of ADMIRE. CONCLUSIONS: The low-contrast detectability of the proposed deep learning-based denoising algorithm was non-inferior to that of ADMIRE and superior to that of FBP. The DLA could successfully improve image quality compared with FBP while showing the similar physical profiles of ADMIRE. KEY POINTS: • Low-contrast detectability in the images denoised using the deep learning algorithm was non-inferior to that in the images reconstructed using standard algorithms. • The proposed deep learning algorithm showed similar profiles of physical measurements to advanced iterative reconstruction algorithm (ADMIRE).

Robustness of Deep Learning Algorithm to Varying Imaging Conditions in Detecting Low Contrast Objects In Computed Tomography Phantom Images: in Comparison to 12 Radiologists.

The performance of deep learning algorithm (DLA) to that of radiologists was compared in detecting low contrast objects in CT phantom images under various imaging conditions. For training, 10,000 images were created using American College of Radiology CT phantom as the background. In half of the images, objects of 3-20 mm size and 5-30 HU contrast difference were generated in random locations. Binary responses were used as the ground truth. For testing, 640 images of Catphan® phantom were used, half of which had objects of either 5 or 9 mm size with 10 HU contrast difference. Twelve radiologists evaluated the presence of objects on a five-point scale. The performances of the DLA and radiologists were compared across different imaging conditions in terms of area under receiver operating characteristics curve (AUC). Multi-reader multi-case AUC and Hanley and McNeil tests were used. We performed post-hoc analysis using bootstrapping and verified that the DLA is less affected by the changing imaging conditions. The AUC of DLA was consistently higher than those of the radiologists across different imaging conditions (p < 0.0001), and it was less affected by varying imaging conditions. The DLA outperformed the radiologists and showed more robust performance under varying imaging conditions.

Deep Learning for Diagnosis of Paranasal Sinusitis Using Multi-View Radiographs.

Accurate image interpretation of Waters' and Caldwell view radiographs used for sinusitis screening is challenging. Therefore, we developed a deep learning algorithm for diagnosing frontal, ethmoid, and maxillary sinusitis on both Waters' and Caldwell views. The datasets were selected for the training and validation set (n = 1403, sinusitis% = 34.3%) and the test set (n = 132, sinusitis% = 29.5%) by temporal separation. The algorithm can simultaneously detect and classify each paranasal sinus using both Waters' and Caldwell views without manual cropping. Single- and multi-view models were compared. Our proposed algorithm satisfactorily diagnosed frontal, ethmoid, and maxillary sinusitis on both Waters' and Caldwell views (area under the curve (AUC), 0.71 (95% confidence interval, 0.62-0.80), 0.78 (0.72-0.85), and 0.88 (0.84-0.92), respectively). The one-sided DeLong's test was used to compare the AUCs, and the Obuchowski-Rockette model was used to pool the AUCs of the radiologists. The algorithm yielded a higher AUC than radiologists for ethmoid and maxillary sinusitis (p = 0.012 and 0.013, respectively). The multi-view model also exhibited a higher AUC than the single Waters' view model for maxillary sinusitis (p = 0.038). Therefore, our algorithm showed diagnostic performances comparable to radiologists and enhanced the value of radiography as a first-line imaging modality in assessing multiple sinusitis.

False-negative results on computer-aided detection software in preoperative automated breast ultrasonography of breast cancer patients.

PURPOSE: The purpose of this study was to measure the cancer detection rate of computer-aided detection (CAD) software in preoperative automated breast ultrasonography (ABUS) of breast cancer patients and to determine the characteristics associated with false-negative outcomes. METHODS: A total of 129 index lesions (median size, 1.7 cm; interquartile range, 1.2 to 2.4 cm) from 129 consecutive patients (mean age±standard deviation, 53.4±11.8 years) who underwent preoperative ABUS from December 2017 to February 2018 were assessed. An index lesion was defined as a breast cancer confirmed by ultrasonography (US)-guided core needle biopsy. The detection rate of the index lesions, positive predictive value (PPV), and false-positive rate (FPR) of the CAD software were measured. Subgroup analysis was performed to identify clinical and US findings associated with false-negative outcomes. RESULTS: The detection rate of the CAD software was 0.84 (109 of 129; 95% confidence interval, 0.77 to 0.90). The PPV and FPR were 0.41 (221 of 544; 95% CI, 0.36 to 0.45) and 0.45 (174 of 387; 95% CI, 0.40 to 0.50), respectively. False-negative outcomes were more frequent in asymptomatic patients (P<0.001) and were associated with the following US findings: smaller size (P=0.001), depth in the posterior third (P=0.002), angular or indistinct margin (P<0.001), and absence of architectural distortion (P<0.001). CONCLUSION: The CAD software showed a promising detection rate of breast cancer. However, radiologists should judge whether CAD software-marked lesions are true- or false-positive lesions, considering its low PPV and high FPR. Moreover, it would be helpful for radiologists to consider the characteristics associated with false-negative outcomes when reading ABUS with CAD.

Diagnostic performance of plain radiography for sacroiliitis in patients with suspected axial spondyloarthritis: a systematic review and meta-analysis.

BACKGROUND: Plain radiography serves a pivotal role in diagnosing axial spondyloarthritis. However, a broad range of diagnostic performance of plain radiography has been reported. PURPOSE: To perform a systematic review and meta-analysis to measure the diagnostic performance of plain radiography for sacroiliitis in patients suspected of having axial spondyloarthritis using magnetic resonance imaging (MRI) findings as the reference standard. MATERIAL AND METHODS: Studies comparing radiography and MRI in the diagnosis of sacroiliitis in patients suspected of having axial spondyloarthritis were searched in PubMed and EMBASE. Additionally, studies analyzed SPondyloaArthritis Caught Early (SPACE), DEvenir des Spondylarthropathies Indifferenciées Récentes (DESIR), GErman Spondyloarthritis Inception Cohort (GESPIC), and South Swedish Arthritis Treatment Group (SSATG) cohorts were manually searched. Pooled sensitivity and specificity of radiography were calculated by using a bivariate random-effects model. Meta-regression analyses were performed to identify the sources of heterogeneity. RESULTS: Eight eligible studies with 1579 patients were included. The pooled sensitivity and specificity of radiography were 0.55 (95% confidence interval [CI] = 0.40-0.69) and 0.87 (95% CI = 0.72-0.95). The meta-regression analyses showed prospective study design and criteria for MRI positivity considering only active bone marrow edema were associated with lower sensitivity. CONCLUSION: The plain radiography showed low sensitivity and reasonable specificity in diagnosis of sacroiliitis in patients suspected of having axial spondyloarthritis.

Comparison of core needle biopsy and fine-needle aspiration in diagnosis of ma lignant salivary gland neoplasm: Systematic review and meta-analysis.

BACKGROUND: In this meta-analysis, we compared the risk of obtaining nondiagnostic results and the diagnostic accuracy for detection of salivary gland malignancy between core needle biopsy (CNB) and fine-needle aspiration (FNA). METHODS: All published English-language studies comparing CNB and FNA diagnostic accuracy for salivary gland masses through December 2019 were searched. Pooled risk ratios (RRs) of nondiagnostic results, sensitivities, and specificities of CNB and FNA for salivary gland malignancy diagnosis were determined. Complication rates were compared. RESULTS: Six studies (1924 procedures) were quantitatively analyzed. CNB yielded significantly fewer nondiagnostic results (P < .001) and had significantly higher pooled sensitivity (P < .001) and specificity (P = .002) than FNA for differentiating malignant and benign salivary gland neoplasms. Hematoma occurred in 0.3% of CNB, while no complication occurred in FNA procedures. CONCLUSION: CNB yielded fewer nondiagnostic results and had superior diagnostic performance compared with FNA for detecting salivary gland malignancies.

Association of meniscal flounce in the knee with the pattern and location of meniscal tear, concomitant ligamentous injury, amount of knee joint effusion, and flexion and rotation angles: a magnetic resonance evaluation.

OBJECTIVE: To determine the association of meniscal flounce with the pattern and location of the meniscal tear, concomitant ligamentous injury, amount of knee joint effusion, and flexion and rotation angles. MATERIALS AND METHODS: A total of 283 knees of 280 patients were retrospectively reviewed over a 9-month period. Thirty-one magnetic resonance images of patients with meniscal flounce were compared with those of age- and sex-matched control group (n = 62) without meniscal flounce. The presence of meniscal tear was evaluated and, if present, its location and pattern were recorded. The amount of joint effusion was graded, and the joint angle was measured. The Fisher's exact, Cochran-Armitage trend, and t tests were performed to compare the findings between the two groups. The decision tree analysis was employed to determine the most significant factor of meniscal flounce. RESULTS: Meniscal flounce was present in 11.0% (31/283) of the adult population. Approximately 80.6% of meniscal flounce occurred in the torn medial menisci. The presence of meniscal flounce was significantly associated with tears at the body (p = 0.007), posterior horn (p = 0.001), and meniscocapsular junction (p = 0.002) of the medial meniscus. The decision tree analysis revealed that the posterior horn tear of the medial meniscus was the most significant predictor of meniscal flounce. CONCLUSION: The most significant factor associated with meniscal flounce is tear at the posterior horn of the medial meniscus, followed by tear at the meniscocapsular junction.

Ruling out rotator cuff tear in shoulder radiograph series using deep learning: redefining the role of conventional radiograph.

OBJECTIVE: To develop a deep learning algorithm that can rule out significant rotator cuff tear based on conventional shoulder radiographs in patients suspected of rotator cuff tear. METHODS: The algorithm was developed using 6793 shoulder radiograph series performed between January 2015 and June 2018, which were labeled based on ultrasound or MRI conducted within 90 days, and clinical information (age, sex, dominant side, history of trauma, degree of pain). The output was the probability of significant rotator cuff tear (supraspinatus/infraspinatus complex tear with > 50% of tendon thickness). An operating point corresponding to sensitivity of 98% was set to achieve high negative predictive value (NPV) and low negative likelihood ratio (LR-). The performance of the algorithm was tested with 1095 radiograph series performed between July and December 2018. Subgroup analysis using Fisher's exact test was performed to identify factors (clinical information, radiography vendor, advanced imaging modality) associated with negative test results and NPV. RESULTS: Sensitivity, NPV, and LR- were 97.3%, 96.6%, and 0.06, respectively. The deep learning algorithm could rule out significant rotator cuff tear in about 30% of patients suspected of rotator cuff tear. The subgroup analysis showed that age < 60 years (p < 0.001), non-dominant side (p < 0.001), absence of trauma history (p = 0.001), and ultrasound examination (p < 0.001) were associated with negative test results. NPVs were higher in patients with age < 60 years (p = 0.024) and examined with ultrasound (p < 0.001). CONCLUSION: The deep learning algorithm could accurately rule out significant rotator cuff tear based on shoulder radiographs. KEY POINTS: • The deep learning algorithm can rule out significant rotator cuff tear with a negative likelihood ratio of 0.06 and a negative predictive value of 96.6%. • The deep learning algorithm can guide patients with significant rotator cuff tear to additional shoulder ultrasound or MRI with a sensitivity of 97.3%. • The deep learning algorithm could rule out significant rotator cuff tear in about 30% of patients with clinically suspected rotator cuff tear.

Resonant inelastic x-ray scattering study of [Formula: see text]-RuCl3: a progress report.

Ru M3-edge resonant inelastic x-ray scattering (RIXS) measurements of [Formula: see text] with 27 meV resolution reveals a spin-orbit exciton without noticeable splitting. We extract values for the spin-orbit coupling constant ([Formula: see text] meV) and trigonal distortion field energy ([Formula: see text] meV) which support the [Formula: see text] nature of [Formula: see text]. We demonstrate the feasibility of M-edge RIXS for 4d systems, which allows ultra high-resolution RIXS of 4d systems until instrumentation for L-edge RIXS improves.

Vav1 inhibits RANKL-induced osteoclast differentiation and bone resorption.

Vav1 is a Rho/Rac guanine nucleotide exchange factor primarily expressed in hematopoietic cells. In this study, we investigated the potential role of Vav1 in osteoclast (OC) differentiation by comparing the ability of bone marrow mononuclear cells (BMMCs) obtained from Vav1-deficient (Vav1-/-) and wild-type (WT) mice to differentiate into mature OCs upon stimulation with macrophage colony stimulating factor and receptor activator of nuclear kappa B ligand in vitro. Our results suggested that Vav1 deficiency promoted the differentiation of BMMCs into OCs, as indicated by the increased expression of tartrate-resistant acid phosphatase, cathepsin K, and calcitonin receptor. Therefore, Vav1 may play a negative role in OC differentiation. This hypothesis was supported by the observation of more OCs in the femurs of Vav1-/- mice than in WT mice. Furthermore, the bone status of Vav1-/- mice was analyzed in situ and the femurs of Vav1-/- mice appeared abnormal, with poor bone density and fewer number of trabeculae. In addition, Vav1-deficient OCs showed stronger adhesion to vitronectin, an αvß3 integrin ligand important in bone resorption. Thus, Vav1 may inhibit OC differentiation and protect against bone resorption. [BMB Reports 2019; 52(11): 659-664].

Clinical and imaging characteristics of patients with extreme low back pain or sciatica referred for spinal injection.

PURPOSE: To analyze the causes of pain, imaging characteristics, and therapeutic effect of spinal injection in patients with extreme low back pain or sciatica. METHODS: We analyzed 381 consecutive patients with extreme low back pain or sciatica visiting our spinal intervention center between January and December 2017. Clinical and imaging characteristics were analyzed. The treatment response, defined as a numerical pain rating scale decrease of ≥ 30%, was measured. Fisher's exact test was performed to identify the association between the injection response and subsequent lumbar surgery rate. RESULTS: The most frequent cause of pain was spinal stenosis, followed by herniated intervertebral disc, facet osteoarthritis, and osteoporotic compression fracture. A herniated intervertebral disc was the most common disorder in patients < 50 years of age, while spinal stenosis was the most common in patients ≥ 50 years of age. Women comprised 66.4% of the study population. The majority of lumbar pathologies occurred below L3/4. Spinal injection was found to be effective in 44.2% of cases. Those who responded to the injection showed a significantly lower rate of lumbar surgery within 6 months (P = 0.004). CONCLUSIONS: Those with extreme low back pain or sciatica had clinical and imaging characteristics similar to those with typical low back pain referred for spinal injection. Spinal injection could be an effective method of pain control for patients with extreme low back pain or sciatica.

Performance of a Deep Learning Algorithm in Detecting Osteonecrosis of the Femoral Head on Digital Radiography: A Comparison With Assessments by Radiologists.

OBJECTIVE. The objective of our study was to compare the sensitivity of a deep learning (DL) algorithm with the assessments by radiologists in diagnosing osteonecrosis of the femoral head (ONFH) using digital radiography. MATERIALS AND METHODS. We performed a two-center, retrospective, noninferiority study of consecutive patients (≥ 16 years old) with a diagnosis of ONFH based on MR images. We investigated the following four datasets of unilaterally cropped hip anteroposterior radiographs: training (n = 1346), internal validation (n = 148), temporal external test (n = 148), and geographic external test (n = 250). Diagnostic performance was measured for a DL algorithm, a less experienced radiologist, and an experienced radiologist. Noninferiority analyses for sensitivity were performed for the DL algorithm and both radiologists. Subgroup analysis for precollapse and postcollapse ONFH was done. RESULTS. Overall, 1892 hips (1037 diseased and 855 normal) were included. Sensitivity and specificity for the temporal external test set were 84.8% and 91.3% for the DL algorithm, 77.6% and 100.0% for the less experienced radiologist, and 82.4% and 100.0% for the experienced radiologist. Sensitivity and specificity for the geographic external test set were 75.2% and 97.2% for the DL algorithm, 77.6% and 75.0% for the less experienced radiologist, and 78.0% and 86.1% for the experienced radiologist. The sensitivity of the DL algorithm was noninferior to that of the assessments by both radiologists. The DL algorithm was more sensitive for precollapse ONFH than the assessment by the less experienced radiologist in the temporal external test set (75.9% vs 57.4%; 95% CI of the difference, 4.5-32.8%). CONCLUSION. The sensitivity of the DL algorithm for diagnosing ONFH using digital radiography was noninferior to that of both less experienced and experienced radiologist assessments.

Deep Learning in Diagnosis of Maxillary Sinusitis Using Conventional Radiography.

OBJECTIVES: The aim of this study was to compare the diagnostic performance of a deep learning algorithm with that of radiologists in diagnosing maxillary sinusitis on Waters' view radiographs. MATERIALS AND METHODS: Among 80,475 Waters' view radiographs, examined between May 2003 and February 2017, 9000 randomly selected cases were classified as normal or maxillary sinusitis based on radiographic findings and divided into training (n = 8000) and validation (n = 1000) sets to develop a deep learning algorithm. Two test sets composed of Waters' view radiographs with concurrent paranasal sinus computed tomography were labeled based on computed tomography findings: one with temporal separation (n = 140) and the other with geographic separation (n = 200) from the training set. Area under the receiver operating characteristics curve (AUC), sensitivity, and specificity of the algorithm and 5 radiologists were assessed. Interobserver agreement between the algorithm and majority decision of the radiologists was measured. The correlation coefficient between the predicted probability of the algorithm and average confidence level of the radiologists was determined. RESULTS: The AUCs of the deep learning algorithm were 0.93 and 0.88 for the temporal and geographic external test sets, respectively. The AUCs of the radiologists were 0.83 to 0.89 for the temporal and 0.75 to 0.84 for the geographic external test sets. The deep learning algorithm showed statistically significantly higher AUC than radiologist in both test sets. In terms of sensitivity and specificity, the deep learning algorithm was comparable to the radiologists. A strong interobserver agreement was noted between the algorithm and radiologists (Cohen κ coefficient, 0.82). The correlation coefficient between the predicted probability of the algorithm and confidence level of radiologists was 0.89 and 0.84 for the 2 test sets, respectively. CONCLUSIONS: The deep learning algorithm could diagnose maxillary sinusitis on Waters' view radiograph with superior AUC and comparable sensitivity and specificity to those of radiologists.