Development of a deep learning method for improving diagnostic accuracy for uterine sarcoma cases.

Uterine sarcomas have very poor prognoses and are sometimes difficult to distinguish from uterine leiomyomas on preoperative examinations. Herein, we investigated whether deep neural network (DNN) models can improve the accuracy of preoperative MRI-based diagnosis in patients with uterine sarcomas. Fifteen sequences of MRI for patients (uterine sarcoma group: n = 63; uterine leiomyoma: n = 200) were used to train the models. Six radiologists (three specialists, three practitioners) interpreted the same images for validation. The most important individual sequences for diagnosis were axial T2-weighted imaging (T2WI), sagittal T2WI, and diffusion-weighted imaging. These sequences also represented the most accurate combination (accuracy: 91.3%), achieving diagnostic ability comparable to that of specialists (accuracy: 88.3%) and superior to that of practitioners (accuracy: 80.1%). Moreover, radiologists' diagnostic accuracy improved when provided with DNN results (specialists: 89.6%; practitioners: 92.3%). Our DNN models are valuable to improve diagnostic accuracy, especially in filling the gap of clinical skills between interpreters. This method can be a universal model for the use of deep learning in the diagnostic imaging of rare tumors.

Deep learning reconstruction for the evaluation of neuroforaminal stenosis using 1.5T cervical spine MRI: comparison with 3T MRI without deep learning reconstruction.

PURPOSE: To compare image quality and interobserver agreement in evaluations of neuroforaminal stenosis between 1.5T cervical spine magnetic resonance imaging (MRI) with deep learning reconstruction (DLR) and 3T MRI without DLR. METHODS: In this prospective study, 21 volunteers (mean age: 42.4 ± 11.9 years; 17 males) underwent cervical spine T2-weighted sagittal 1.5T and 3T MRI on the same day. The 1.5T and 3T MRI data were used to reconstruct images with (1.5T-DLR) and without (3T-nonDLR) DLR, respectively. Regions of interest were marked on the spinal cord to calculate non-uniformity (NU; standard deviation/signal intensity × 100), as an indicator of image noise. Two blinded radiologists evaluated the images in terms of the depiction of structures, artifacts, noise, overall image quality, and neuroforaminal stenosis. The NU value and the subjective image quality scores were compared between 1.5T-DLR and 3T-nonDLR using the Wilcoxon signed-rank test. Interobserver agreement in evaluations of neuroforaminal stenosis for 1.5T-DLR and 3T-nonDLR was evaluated using Cohen's weighted kappa analysis. RESULTS: The NU value for 1.5T-DLR was 8.4, which was significantly better than that for 3T-nonDLR (10.3; p < 0.001). Subjective image scores were significantly better for 1.5T-DLR than 3T-nonDLR images (p < 0.037). Interobserver agreement (95% confidence intervals) in the evaluations of neuroforaminal stenosis was significantly superior for 1.5T-DLR (0.920 [0.916-0.924]) than 3T-nonDLR (0.894 [0.889-0.898]). CONCLUSION: By using DLR, image quality and interobserver agreement in evaluations of neuroforaminal stenosis on 1.5T cervical spine MRI could be improved compared to 3T MRI without DLR.

Deep learning reconstruction for 1.5 T cervical spine MRI: effect on interobserver agreement in the evaluation of degenerative changes.

OBJECTIVES: To investigate whether deep learning reconstruction (DLR) provides improved cervical spine MR images using a 1.5 T unit in the evaluation of degenerative changes without increasing imaging time. METHODS: This study included 21 volunteers (age 42.4 ± 11.9 years; 17 males) who underwent 1.5 T cervical spine sagittal T2-weighted MRI. From the imaging data with number of acquisitions (NAQ) of 1 or 2, images were reconstructed with DLR (NAQ1-DLR) and without DLR (NAQ1) or without DLR (NAQ2), respectively. Two readers evaluated the images for depiction of structures, artifacts, noise, overall image quality, spinal canal stenosis, and neuroforaminal stenosis. The two readers read studies blinded and randomly. Values were compared between NAQ1-DLR and NAQ1 and between NAQ1-DLR and NAQ2 using the Wilcoxon signed-rank test. RESULTS: The analyses showed significantly better results for NAQ1-DLR compared with NAQ1 and NAQ2 (p < 0.023), except for depiction of disc and foramina by one reader and artifacts by both readers in the comparison between NAQ1-DLR and NAQ2. Interobserver agreements (Cohen's weighted kappa [97.5% confidence interval]) in the evaluation of spinal canal stenosis for NAQ1-DLR/NAQ1/NAQ2 were 0.874 (0.866-0.883)/0.778 (0.767-0.789)/0.818 (0.809-0.827), respectively, and those in the evaluation of neuroforaminal stenosis were 0.878 (0.872-0.883)/0.855 (0.849-0.860)/0.852 (0.845-0.860), respectively. CONCLUSIONS: DLR improved the 1.5 T cervical spine MR images in the evaluation of degenerative spine changes. KEY POINTS: • Two radiologists demonstrated that deep learning reconstruction reduced the noise in cervical spine sagittal T2-weighted MR images obtained using a 1.5 T unit. • Reduced noise in deep learning reconstruction images resulted in a clearer depiction of structures, such as the spinal cord, vertebrae, and zygapophyseal joint. • Interobserver agreement in the evaluation of spinal canal stenosis and foraminal stenosis on cervical spine MR images was significantly improved using deep learning reconstruction (0.874 and 0.878, respectively) versus without deep learning (0.778-0.818 and 0.852-0.855, respectively).

Incidence and computed tomography findings of lenvatinib-induced pancreatobiliary inflammation: A single-center, retrospective study.

ABSTRACT: In this single-center retrospective study, we intended to evaluate the frequencies and characteristics of computed tomography findings of pancreatobiliary inflammation (PBI) in patients treated with lenvatinib and the relationship of these findings with treatment-planning changes.We included 78 patients (mean ±â€Šstandard deviation, 69.8 ±â€Š9.4 years, range: 39-84 years, 62 men) with hepatocellular carcinoma (n = 62) or thyroid carcinoma (n = 16) who received lenvatinib (June 2016-September 2020). Two radiologists interpreted the posttreatment computed tomography images and assessed the radiological findings of PBI (symptomatic pancreatitis, cholecystitis, or cholangitis). The PBI effect on treatment was statistically evaluated.PBI (pancreatitis, n = 1; cholecystitis, n = 7; and cholangitis, n = 2) was diagnosed in 11.5% (9/78) of the patients at a median of 35 days after treatment initiation; 6 of 9 patients discontinued treatment because of PBI. Three cases of cholecystitis and 1 of cholangitis were accompanied by gallstones, while the other 5 were acalculous. The treatment duration was significantly shorter in patients with PBI than in those without (median: 44 days vs. 201 days, P = .02). Overall, 9 of 69 patients without PBI showed asymptomatic gallbladder subserosal edema.Lenvatinib-induced PBI developed in 11.5% of patients, leading to a significantly shorter treatment duration. Approximately 55.6% of the PBI cases were acalculous. The recognition of this phenomenon would aid physicians during treatment planning in the future.

CT imaging findings of lenvatinib-induced enteritis.

OBJECTIVES: To evaluate the relationship between abnormal findings on abdomino-pelvic CT and adverse events in oncologic patients treated with lenvatinib, and their relationship with treatment planning. METHODS: This single institutional retrospective study included 58 patients with unresectable hepatocellular carcinoma or unresectable thyroid carcinoma (mean age ± standard deviation 69.6 ± 10.0 years; range 39-84 years; 48 men) who underwent CT between October 2016 and July 2020. Two radiologists who were blinded to clinical information including the presence or absence of diarrhea evaluated the imaging findings, including the presence/absence of enteritis in each intestinal segment. Gastrointestinal adverse events (diarrhea, decreased appetite, nausea, and vomiting) and other drug-induced adverse events requiring treatment or follow-up during lenvatinib treatment were also investigated. The frequency of these adverse events was compared between the patients with and without enteritis using Fisher's exact test or the Mann-Whitney U test. RESULTS: Enteritis was found on CT in the majority (33/58 [56.9%]) of the patients, and most of them (25/33 [75.8%]) showed duodenojejunitis. The frequency of gastrointestinal adverse events (28/33 [84.8%] vs. 13/25 [56.0%], p = 0.009), diarrhea (20/33 [60.6%] vs. 3/25 [12.0%], p < 0.001), and drug interruptions (25/33 [75.8%] vs. 10/25 [40.0%], p = 0.008) and the number of other adverse events (3.9 ± 1.7 vs. 2.3 ± 1.3, p < 0.001) were significantly higher in the patients with enteritis on CT than in those without. CONCLUSIONS: Lenvatinib-induced enteritis frequently involved the duodenum and jejunum and was related to a significantly higher frequency of treatment interruptions and gastrointestinal adverse events.

Forward-projected Model-based Iterative Reconstruction SoluTion in Temporal Bone Computed Tomography: A Comparison Study of All Reconstruction Modes.

OBJECTIVE: Identify appropriate reconstruction modes of Forward-projected model-based Iterative Reconstruction SoluTion (FIRST) in temporal bone computed tomography (CT) and investigate the contribution of spatial resolution and noise to the visual assessment. METHODS: Sixteen temporal bone CT images were reconstructed. Two blinded radiologists assessed 8 anatomical structures and classified the visual assessment. These visual scores were compared across reconstruction modes among each anatomical structure. Visual scores and contrast-to-noise ratio, noise power spectrum (NPS) at low (NPSLow) and high (NPSHigh) spatial frequencies, and 50% modulation transfer function relationships were evaluated. RESULTS: Visual scores differed significantly for the stapedius muscle and osseous spiral lamina, with CARDIAC SHARP, BONE, and LUNG modes for the osseous spiral lamina scoring highest. Visual scores significantly positively correlated with NPSLow, NPSHigh, and 50% modulation transfer function but negatively with contrast-to-noise ratio. CONCLUSIONS: Modes providing higher spatial resolution and lower noise reduction showed an improved visual assessment of CT images reconstructed with FIRST.