18F-FDG PET/CT Uptake Classification in Lymphoma and Lung Cancer by Using Deep Convolutional Neural Networks.

Background Fluorine 18 (18F)-fluorodeoxyglucose (FDG) PET/CT is a routine tool for staging patients with lymphoma and lung cancer. Purpose To evaluate configurations of deep convolutional neural networks (CNNs) to localize and classify uptake patterns of whole-body 18F-FDG PET/CT images in patients with lung cancer and lymphoma. Materials and Methods This was a retrospective analysis of consecutive patients with lung cancer or lymphoma referred to a single center from August 2011 to August 2013. Two nuclear medicine experts manually delineated foci with increased 18F-FDG uptake, specified the anatomic location, and classified these findings as suspicious for tumor or metastasis or nonsuspicious. By using these expert readings as the reference standard, a CNN was developed to detect foci positive for 18F-FDG uptake, predict the anatomic location, and determine the expert classification. Examinations were divided into independent training (60%), validation (20%), and test (20%) subsets. Results This study included 629 patients (mean age, 52.2 years ± 20.4 [standard deviation]; 394 men). There were 302 patients with lung cancer and 327 patients with lymphoma. For the test set (123 patients; 10 782 foci), the CNN areas under the receiver operating characteristic curve (AUCs) for determining hypermetabolic 18F-FDG PET/CT foci that were suspicious for cancer versus nonsuspicious by using the five input features were as follows: CT alone, 0.78 (95% confidence interval [CI]: 0.72, 0.83); 18F-FDG PET alone, 0.97 (95% CI: 0.97, 0.98); 18F-FDG PET/CT, 0.98 (95% CI: 0.97, 0.99); 18F-FDG PET/CT maximum intensity projection (MIP), 0.98 (95% CI: 0.98, 0.99); and 18F-FDG PET/CT MIP atlas, 0.99 (95% CI: 0.98, 1.00). The combination of 18F-FDG PET and CT information improved overall classification accuracy (AUC, 0.975 vs 0.981, respectively; P < .001). Anatomic localization accuracy of the CNN was 2543 of 2639 (96.4%; 95% CI: 95.5%, 97.1%) for body part, 2292 of 2639 (86.9%; 95% CI: 85.3%, 88.5%) for region (ie, organ), and 2149 of 2639 (81.4%; 95% CI: 79.3%-83.5%) for subregion. Conclusion The fully automated anatomic localization and classification of fluorine 18-fluorodeoxyglucose PET uptake patterns in foci suspicious and nonsuspicious for cancer in patients with lung cancer and lymphoma by using a convolutional neural network is feasible and achieves high diagnostic performance when both CT and PET images are used. © RSNA, 2019 Online supplemental material is available for this article. See also the editorial by Froelich and Salavati in this issue.

Impact of Data-driven Respiratory Gating in Clinical PET.

Purpose To study the feasibility and impact of respiratory gating in positron emission tomographic (PET) imaging in a clinical trial comparing conventional hardware-based gating with a data-driven approach and to describe the distribution of determined parameters. Materials and Methods This prospective study was approved by the ethics committee of the University Hospital of Münster (AZ 2014-217-f-N). Seventy-four patients suspected of having abdominal or thoracic fluorine 18 fluorodeoxyglucose (FDG)-positive lesions underwent clinical whole-body FDG PET/computed tomographic (CT) examinations. Respiratory gating was performed by using a pressure-sensitive belt system (belt gating [BG]) and an automatic data-driven approach (data-driven gating [DDG]). PET images were analyzed for lesion uptake, metabolic volumes, respiratory shifts of lesions, and diagnostic image quality. Results Forty-eight patients had at least one lesion in the field of view, resulting in a total of 164 lesions analyzed (range of number of lesions per patient, one to 13). Both gating methods revealed respiratory shifts of lesions (4.4 mm ± 3.1 for BG vs 4.8 mm ± 3.6 for DDG, P = .76). Increase in uptake of the lesions compared with nongated values did not differ significantly between both methods (maximum standardized uptake value [SUVmax], +7% ± 13 for BG vs +8% ± 16 for DDG, P = .76). Similarly, gating significantly decreased metabolic lesion volumes with both methods (-6% ± 26 for BG vs -7% ± 21 for DDG, P = .44) compared with nongated reconstructions. Blinded reading revealed significant improvements in diagnostic image quality when using gating, without significant differences between the methods (DDG was judged to be inferior to BG in 22 cases, equal in 12 cases, and superior in 15 cases; P = .32). Conclusion Respiratory gating increases diagnostic image quality and uptake values and decreases metabolic volumes compared with nongated acquisitions. Data-driven approaches are clinically applicable alternatives to belt-based methods and might help establishing routine respiratory gating in clinical PET/CT. (©) RSNA, 2016 Online supplemental material is available for this article.

Cross-sectional validation of the Rasch-based Depression Screening (DESC) in a mixed sample of patients with mental and somatic diseases.

BACKGROUND: The study aimed to cross-validate the psychometric properties of the two parallel versions of the "Rasch-based Depression Screening (DESC)" in a mixed clinical sample of patients with mental disorders and somatic diseases. Additionally, it was intended to confirm the initially proposed cut-off scores. METHODS: One hundred eleven inpatients from the Departments of Psychiatry (n=50), Cardiology (n=39) and Otorhinolaryngology (n=22) were examined. Accordance to Rasch model assumptions was determined by analyzing item infit and outfit as well as unidimensionality and local independence via residual principal components factor analysis (PCFA). Furthermore, separation, reliability and the targeting were considered. Differential item functioning (DIF) was investigated with regard to gender, age and clinical group. Finally, receiver operating characteristics (ROC) curves analyses were conducted and parallel test reliability was calculated. RESULTS: Item and person separation and reliability were good with separation values being above 2.90 and reliability exceeding being .90. Rasch model fit was good for all items except for the two items "sad" and "hopeless" displaying slight misfit. Targeting was appropriate with a person mean of -1.6 (DESC-I) and -1.5 (DESC-II) reflecting as expected that the two DESC versions were in average to easy for a mixed sample of patients with mental disorders and somatic. PCFA showed unidimensionality and local independence: The Rasch dimension "depression" explained 84.8% (DESC-I) and 82.2% (DESC-II) of the observed variance. No DIF was found. ROC curves confirmed the proposed cut-off score of 12 points. The area under the curve amounted to .91 for both DESC forms. The Youden index was 75.5 for DESC-I and 75.7 for DESC-II. A high parallel test reliability of .94 showed virtual equivalence of both DESC forms. CONCLUSIONS: Results of the present study support the good psychometric properties of both DESC forms in patients with mental disorders and somatic diseases. This study confirmed the cut-off scores that had already been determined in former patient samples. Thus, both DESC forms can be used validly as screening instruments for depression in clinical practice.

Prevalence and psychopathological characteristics of depression in consecutive otorhinolaryngologic inpatients.

BACKGROUND: High prevalence of depression has been reported in otorhinolaryngologic patients (ORL). However, studies using a semi-structured interview to determine the prevalence of depression in ORL are lacking. Therefore the present study sought to determine the depression prevalence in ORL applying a semi-structured diagnostic interview and to further characterize the pathopsychological and demographic characteristics of depression in these patients. METHODS: One-hundred inpatients of the otorhinolaryngologic department of a German university hospital participated voluntarily (age M = 38.8 years, SD = 13.9; 38.0% female). Depression was assessed using a clinical interview in which the International Diagnostic Checklist for depression (IDCL) was applied. Patients completed the Brief Symptom Inventory (BSI) which constitutes three composite scores and nine symptom scales and the Beck Depression Inventory (BDI). Multivariate analyses of variance, correlations and effect sizes were conducted. RESULTS: A prevalence of depression of 21.0% was determined, 38.0% of the depressed patients were female. Depressed patients showed higher scores on the BSI-scales "interpersonal sensitivity", "depression", "anxiety", "phobic anxiety" and "psychoticism" with medium effect sizes. CONCLUSIONS: High prevalence of depression was found which is in accordance with results of prior studies. Depressed patients showed higher psychological distress as compared to non-depressed patients. The results call for carrying on in engaging in depression research and routine depression screening in ORL.

Sensitivity and specificity of the Beck Depression Inventory in cardiologic inpatients: how useful is the conventional cut-off score?

OBJECTIVE: The Beck Depression Inventory (BDI) is widely used for depression screening in various patient populations. However, there are still insufficient data about its sensitivity and specificity in nonpsychiatric patients. Furthermore, some research suggests that somatic BDI items heighten its sum score artificially in physically ill patients. The aim of the present study was to validate the conventional BDI cut-off score by examination of its sensitivity and specificity in a mixed sample of cardiac inpatients and compare it to a modified "cognitive-emotional" BDI (BDI(c/e)) after exclusion of somatic items. METHODS: A total of 126 cardiologic inpatients were assessed. Receiver operating characteristic curves (ROC) were calculated for total BDI (BDI(t)) and BDI(c/e). Screening performance of cut-off scores was evaluated using the Youden Index (Y). RESULTS: With the application of the conventional BDI cut-off score, ROC analysis revealed a moderate overall screening performance with Y=52.6 and an area under the curve (AUC) of 0.83. In contrast, Y improved to 57.5 at a cut-off score of >9, but screening performance was still not optimal. BDI(c/e) showed also a moderate screening performance (AUC=.82); Y was maximized at a cut-off score of >8 (Y=0.53.5). Again, no cut-off score provided optimal screening performance. CONCLUSION: The BDI cannot be recommended as a formal screening instrument in cardiac inpatients since no cut-off score for either BDI(t) or BDI(c/e) combined both sufficiently high sensitivity and specificity. However, the shorter BDI(c/e) could be used as alternative to BDI(t) which may be confounded in physically ill patients. Generally, researchers should consider using alternative screening instruments (e.g., the Hospital Anxiety and Depression Scale) instead.