RESUMO
OBJECTIVE: The purpose of this pilot study was to examine human and automated estimates of reporting complexity for computed tomography (CT) studies of the abdomen and pelvis. METHODS: A total of 1019 CT studies were reviewed and categorized into 3 complexity categories by 3 abdominal radiologists, and the majority classification was used as ground truth. Studies were randomized into a training set of 498 studies and a test set of 521 studies. A 2-stage neural network model was trained on the training set; the first-stage image-level classifier produces image embeddings that are used in the second-stage sequential model to provide a study-level prediction. RESULTS: All 3 human reviewers agreed on ratings for 470 of the 1019 studies (46%); at least 2 of the 3 reviewers agreed on ratings for 1010 studies (99%). After training, the neural network model predicted complexity labels that agreed with the radiologist consensus rating on 55% of the studies; 90% of the incorrect predicted categories were errors where the predicted category differed from the consensus rating by one level of complexity. CONCLUSIONS: There is moderate interrater agreement in radiologist-perceived reporting complexity for CT studies of the abdomen and pelvis. Automated prediction of reporting complexity in radiology studies may be a useful adjunct to radiology practice analytics.
Assuntos
Aprendizado Profundo , Humanos , Redes Neurais de Computação , Pelve/diagnóstico por imagem , Projetos Piloto , Tomografia Computadorizada por Raios X/métodosRESUMO
BACKGROUND: Sarcopenia is associated with postoperative complications in inflammatory bowel disease. It has most commonly been defined using the skeletal muscle index, computed after analysis of cross-sectional muscle area at L3. Psoas muscle thickness normalized to height (PMTH), which is easier to derive, is a potential surrogate of SMI and sarcopenia in patients with cirrhosis and chronic pancreatitis. We investigate whether sarcopenia defined by PMTH has utility in predicting post-operative outcomes in patients with inflammatory bowel disease. METHODS: We performed a retrospective study of adults undergoing IBD-related surgery from 2009 to 2019 at two hospitals. Sarcopenia was defined by sex-specific PMTH at the umbilicus on cross-sectional imaging using a 50th percentile median cutoff. Predictive models were created using variables (BMI, age, sex, smoking status, albumin, INR, platelets, hemoglobin, hypertension, diabetes, CAD, medications) that may be associated with complications (mortality, reoperation, readmission, transfusions, ICU admission, infection, DVT/PE), and sarcopenia for comparison. RESULTS: 85 patients with IBD were included. Lower albumin level (OR = 0.52, p = 0.039) and biologic use (OR = 5.92, p = 0.006) were associated with postoperative complications. There was no significant difference using PMTH compared to a model incorporating hypoalbuminemia and biologic use in predicting complications. Sarcopenia on univariate analysis was associated with a lower 30 day rate of reoperation (p = 0.04). CONCLUSIONS: A low status of PMTH was not associated with increased postoperative complications, however hypoalbuminemia and biologic use were. PMTH as a surrogate for sarcopenia requires further study, ideally with prospective studies comparing PMTH with accepted radiographic surrogates for sarcopenia, to determine its role in clinical decision making.
Assuntos
Doenças Inflamatórias Intestinais , Sarcopenia , Adulto , Feminino , Humanos , Doenças Inflamatórias Intestinais/patologia , Masculino , Complicações Pós-Operatórias , Estudos Prospectivos , Músculos Psoas/diagnóstico por imagem , Músculos Psoas/patologia , Estudos Retrospectivos , Sarcopenia/complicações , Sarcopenia/diagnóstico por imagem , Tomografia Computadorizada por Raios XRESUMO
OBJECTIVE: The purpose of this study was to evaluate improvement of convolutional neural network detection of high-grade small-bowel obstruction on conventional radiographs with increased training set size. MATERIALS AND METHODS: A set of 2210 abdominal radiographs from one institution (image set 1) had been previously classified into obstructive and nonobstructive categories by consensus judgments of three abdominal radiologists. The images were used to fine-tune an initial convolutional neural network classifier (stage 1). An additional set of 13,935 clinical images from the same institution was reduced to 5558 radiographs (image set 2) primarily by retaining only images classified positive for bowel obstruction by the initial classifier. These images were classified into obstructive and nonobstructive categories by an abdominal radiologist. The combined 7768 radiographs were used to train additional classifiers (stage 2 training). The best classifiers from stage 1 and stage 2 training were evaluated on a held-out test set of 1453 abdominal radiographs from image set 1. RESULTS: The ROC AUC for the neural network trained on image set 1 was 0.803; after stage 2, the ROC AUC of the best model was 0.971. By use of an operating point based on maximizing the validation set Youden J index, the stage 2-trained model had a test set sensitivity of 91.4% and specificity of 91.9%. Classification performance increased with training set size, reaching a plateau with over 200 positive training examples. CONCLUSION: Accuracy of detection of high-grade small-bowel obstruction with a convolutional neural network improves significantly with the number of positive training radiographs.
Assuntos
Obstrução Intestinal/diagnóstico por imagem , Intestino Delgado/diagnóstico por imagem , Redes Neurais de Computação , Adolescente , Adulto , Idoso , Idoso de 80 Anos ou mais , Feminino , Humanos , Masculino , Pessoa de Meia-Idade , Radiografia/métodos , Estudos Retrospectivos , Adulto JovemRESUMO
OBJECTIVE. The objective of this study was to determine if moderate to severe tricuspid regurgitation (TR) can be predicted on routine non-ECG-gated CT images of the chest with contrast agent. MATERIALS AND METHODS. Non-ECG-gated CT images of the chest in 674 people who had undergone echocardiography within 24 hours of CT were retrospectively reviewed. CT images were reviewed, and measurements of the tricuspid annular diameter were recorded. Echocardiogram reports were independently reviewed, and the presence and severity of TR was recorded. TR was graded in a multiparametric approach using a combination of qualitative, quantitative, and objective measures. TR grades of either "moderate" or "severe" were considered significant, and grades reported as "no regurgitation," "trace," "trivial," or "mild" were considered insignificant. RESULTS. The overall prevalence of significant (moderate or severe) TR was 15.0% (10.7% prevalence in men and 19.8% in women). Using ≥ 41 mm as a cut point on CT, the prediction accuracy reached an AUC of 0.92 (95% CI, 0.89-0.96). Specificity was 96.5% (95% CI, 95.0-97.9%), sensitivity was 75.7% (95% CI, 67.6-83.8%), positive predictive value was 80.2%, and negative predictive value was 95.5%. CONCLUSION. Using a cut point of ≥ 41 mm, measurement of the tricuspid valve annulus on routine CT of the chest can predict moderate to severe TR with a specificity of 96.5% and sensitivity of 75.7%. Measurements below threshold had a 95.5% negative predictive value. Tricuspid annular diameter can be measured on routine contrast-enhanced CT images of the chest, and, when above threshold, further evaluation with echocardiography should be recommended.
Assuntos
Tomografia Computadorizada por Raios X , Insuficiência da Valva Tricúspide/diagnóstico por imagem , Valva Tricúspide/diagnóstico por imagem , Idoso , Meios de Contraste , Ecocardiografia , Feminino , Humanos , Masculino , Pessoa de Meia-Idade , Valor Preditivo dos Testes , Estudos Retrospectivos , Índice de Gravidade de DoençaRESUMO
The American Society of Emergency Radiology 2012 Annual Scientific Meeting and Post-Graduate Course encompassed a wide range of topics: traumatic and non-traumatic emergencies, radiation dose management, technical innovations and advancements, emergency preparedness, mass casualties, and teleradiology. This article highlights the scientific and educational abstracts presented at the meeting.
Assuntos
Diagnóstico por Imagem , Medicina de Emergência , Ferimentos e Lesões/diagnóstico , Congressos como Assunto , Humanos , Sociedades Médicas , Estados UnidosRESUMO
The purpose of this pilot study is to determine whether a deep convolutional neural network can be trained with limited image data to detect high-grade small bowel obstruction patterns on supine abdominal radiographs. Grayscale images from 3663 clinical supine abdominal radiographs were categorized into obstructive and non-obstructive categories independently by three abdominal radiologists, and the majority classification was used as ground truth; 74 images were found to be consistent with small bowel obstruction. Images were rescaled and randomized, with 2210 images constituting the training set (39 with small bowel obstruction) and 1453 images constituting the test set (35 with small bowel obstruction). Weight parameters for the final classification layer of the Inception v3 convolutional neural network, previously trained on the 2014 Large Scale Visual Recognition Challenge dataset, were retrained on the training set. After training, the neural network achieved an AUC of 0.84 on the test set (95% CI 0.78-0.89). At the maximum Youden index (sensitivity + specificity-1), the sensitivity of the system for small bowel obstruction is 83.8%, with a specificity of 68.1%. The results demonstrate that transfer learning with convolutional neural networks, even with limited training data, may be used to train a detector for high-grade small bowel obstruction gas patterns on supine radiographs.
Assuntos
Interpretação de Imagem Assistida por Computador/métodos , Obstrução Intestinal/diagnóstico por imagem , Intestino Delgado/diagnóstico por imagem , Redes Neurais de Computação , Radiografia/métodos , Adolescente , Adulto , Idoso , Idoso de 80 Anos ou mais , Feminino , Humanos , Masculino , Pessoa de Meia-Idade , Projetos Piloto , Sensibilidade e Especificidade , Adulto JovemRESUMO
PURPOSE: To identify long-term post-operative imaging findings resulting from right lobe liver donation. METHODS: This retrospective imaging review consisted of 42 adults (20 males, 22 females, mean age 36.6 years, range 18.7-55.9 years) who underwent right lobe liver donation with pre- and post-operative imaging between 1999 and 2006. The mean follow-up period was 30.2 months (range 12.3-69.6 months). Follow-up imaging evaluation included assessment of the biliary tree, particularly isolated bile ducts draining to and terminating at the cut surface (orphan ducts). Three-dimensional volumetry of the liver remnant was also assessed. RESULTS: After liver donation, 29/42 (69%) participants demonstrated orphan ducts on follow-up imaging (95% confidence interval 52.9% to 82.4%). In those patients with orphan ducts, the main draining ducts were normal in 70% and dilated in 69%. Nearly all right lobe liver donors with orphan ducts had no clinical symptoms on follow-up (28/29); the only liver donor with clinical symptoms on follow-up was subsequently diagnosed with primary biliary cirrhosis. Mean regenerated liver volume was approximately 93%, with mean pre-operative total liver volume of 1552 ml (median 1504 ml, range 1040-2520 ml) and mean post-operative total liver volume of 1446 ml (median 1455 ml, range 964-2090 ml). CONCLUSIONS: Orphan ducts are changes that may be seen after liver donation. The presence of these findings in the absence of clinical symptoms or abnormal hepatic chemistries does not require further work-up and should not be considered pathologic.