Thomas Frauenfelder

Prof. Thomas Frauenfelder

Institute of diagn. and interv. Radiology, University Hospital of Zurich, Zürich, Switzerland
orcid

Research Focus & Metrics

Academic Career Length

24Years (2003-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
19 September 2006 in European Radiology
Verified
Journal Article
Open access
PDF file
HSHans Scheffel
Hans Scheffel
Hatem AlkadhiHatem Alkadhi
Hatem Alkadhi
APAndré Plass
André Plass
RVRobert Vachenauer
Robert Vachenauer
LDLotus Desbiolles
Lotus Desbiolles
SLSebastian Leschka
Sebastian Leschka
The aim of this study was to assess the diagnostic accuracy of dual-source computed tomography (DSCT) for evaluation of coronary artery disease (CAD) in a population with extensive coronary calcifications without heart rate control. Thirty patients (24 male, 6 female, mean age 63.1±11.3 years) with a high pre-test probability of CAD underwent DSCT coronary angiography and invasive coronary angiography (ICA) within 14±9 days. No beta-blockers were administered prior to the scan. Two readers independently assessed image quality of all coronary segments with a diameter ≥1.5 mm using a four-point score (1: excellent to 4: not assessable) and qualitatively assessed significant stenoses as narrowing of the luminal diameter >50%. Causes of false-positive (FP) and false-negative (FN) ratings were assigned to calcifications or motion artifacts. ICA was considered the standard of reference. Mean body mass index was 28.3±3.9 kg/m2 (range 22.4–36.3 kg/m2), mean heart rate during CT was 70.3±14.2 bpm (range 47–102 bpm), and mean Agatston score was 821±904 (range 0–3,110). Image quality was diagnostic (scores 1–3) in 98.6% (414/420) of segments (mean image quality score 1.68±0.75); six segments in three patients were considered not assessable (1.4%). DSCT correctly identified 54 of 56 significant coronary stenoses. Severe calcifications accounted for false ratings in nine segments (eight FP/one FN) and motion artifacts in two segments (one FP/one FN). Overall sensitivity, specificity, positive and negative predictive value for evaluating CAD were 96.4, 97.5, 85.7, and 99.4%, respectively. First experience indicates that DSCT coronary angiography provides high diagnostic accuracy for assessment of CAD in a high pre-test probability population with extensive coronary calcifications and without heart rate control.

1 July 2017 in Investigative Radiology
Verified
Journal Article
Closed access
PDF file
ABAnton S. Becker
Anton S. Becker
MMMagda Marcon
Magda Marcon
SGSoleen Ghafoor
Soleen Ghafoor
MWMoritz C. Wurnig
Moritz C. Wurnig
Thomas FrauenfelderThomas Frauenfelder
Thomas Frauenfelder
ABAndreas Boss
Andreas Boss
Objectives: The aim of this study was to evaluate the diagnostic accuracy of a multipurpose image analysis software based on deep learning with artificial neural networks for the detection of breast cancer in an independent, dual-center mammography data set. Materials and Methods: In this retrospective, Health Insurance Portability and Accountability Act-compliant study, all patients undergoing mammography in 2012 at our institution were reviewed (n = 3228). All of their prior and follow-up mammographies from a time span of 7 years (2008–2015) were considered as a reference for clinical diagnosis. After applying exclusion criteria (missing reference standard, prior procedures or therapies), patients with the first diagnosis of a malignoma or borderline lesion were selected (n = 143). Histology or clinical long-term follow-up served as reference standard. In a first step, a breast density-and age-matched control cohort was selected (n = 143) from the remaining patients with more than 2 years follow-up (n = 1003). The neural network was trained with this data set. From the publicly available Breast Cancer Digital Repository data set, patients with cancer and a matched control cohort were selected (n = 35 × 2). The performance of the trained neural network was also tested with this external data set. Three radiologists (3, 5, and 10 years of experience) evaluated the test data set. In a second step, the neural network was trained with all cases from January to September and tested with cases from October to December 2012 (screening-like cohort). The radiologists also evaluated this second test data set. The areas under the receiver operating characteristic curve between readers and the neural network were compared. A Bonferroni-corrected P value of less than 0.016 was considered statistically significant. Results: Mean age of patients with lesion was 59.6 years (range, 35–88 years) and in controls, 59.1 years (35–83 years). Breast density distribution (A/B/C/D) was 21/59/42/21 and 22/60/41/20, respectively. Histologic diagnoses were invasive ductal carcinoma in 90, ductal in situ carcinoma in 13, invasive lobular carcinoma in 13, mucinous carcinoma in 3, and borderline lesion in 12 patients. In the first step, the area under the receiver operating characteristic curve of the trained neural network was 0.81 and comparable on the test cases 0.79 (P = 0.63). One of the radiologists showed almost equal performance (0.83, P = 0.17), whereas 2 were significantly better (0.91 and 0.94, P < 0.016). In the second step, performance of the neural network (0.82) was not significantly different from the human performance (0.77–0.87, P > 0.016); however, radiologists were consistently less sensitive and more specific than the neural network. Conclusions: Current state-of-the-art artificial neural networks for general image analysis are able to detect cancer in mammographies with similar accuracy to radiologists, even in a screening-like cohort with low breast cancer prevalence.

1 July 2014 in Investigative Radiology
Verified
Journal Article
Closed access
SGSonja Gordic
Sonja Gordic
FMFabian Morsbach
Fabian Morsbach
BSBernhard Schmidt
Bernhard Schmidt
TAThomas Allmendinger
Thomas Allmendinger
TFThomas Flohr
Thomas Flohr
HAHatem Alkadhi
Hatem Alkadhi
The purpose of this study was to evaluate the image quality and sensitivity of ultralow radiation dose single-energy computed tomography (CT) with tin filtration for spectral shaping and iterative reconstructions for the detection of pulmonary nodules in a phantom setting. Single-energy CT was performed using third-generation dual-source CT (SOMATOM Force; 2 × 192 slices) at 70 kVp, 100 kVp with tin filtration (100Sn kVp), and 150Sn kV with tube current-time product adjustments resulting in standard dose (CT volume dose index, 3.1 mGy/effective dose, 1.3 mSv at a scan length of 30 cm), 1/10th dose level (0.3 mGy/0.13 mSv), and 1/20th dose level (0.15 mGy/0.06 mSv). An anthropomorphic chest phantom simulating an intermediate-sized adult with randomly distributed solid pulmonary nodules of various sizes (2-10 mm; attenuation, 75 HU at 120 kVp) was used. Images were reconstructed with advanced model-based iterative reconstruction (ADMIRE; strength levels 3 and 5) and were compared with those acquired with second-generation dual-source CT at 120 kVp (reconstructed with filtered back projection) and sinogram-affirmed iterative reconstruction (strength level 3) at the lowest possible dose at 120 kVp (CT volume dose index, 0.28 mGy). One blinded reader measured image noise, and 2 blinded, independent readers determined overall image quality on a 5-grade scale (1 = nondiagnostic to 5 = excellent) and marked nodule localization with confidence rates on a 5-grade scale (1 = unsure to 5 = high confidence). The constructional drawing of the phantom served as reference standard for calculation of sensitivity. Two patients were included, for proof of concept, who were scanned with the 100Sn kVp protocol at the 1/10th and 1/20th dose level. Image noise was highest in the images acquired with second-generation dual-source CT and reconstructed with filtered back projection. At both the 1/10th and 1/20th dose levels, image noise at a tube voltage of 100Sn kVp was significantly lower than in the 70 kVp and 150Sn kV data sets (ADMIRE 3, P < 0.01; ADMIRE 5, P < 0.05). Sensitivity of nodule detection was lowest in images acquired with second-generation dual-source CT at 120 kVp and the lowest possible dose. Protocols at 100Sn kVp and ADMIRE 5 showed highest sensitivity at the 1/10th and 1/20th dose levels. Highest numbers of false-positives occurred in second-generation dual-source CT images (range, 12-15), whereas lowest numbers occurred in the 1/10th and 1/20th dose data sets acquired with third-generation dual-source CT at 100Sn kVp and reconstructed with ADMIRE strength levels 3 and 5 (total of 1 and 0 false-positives, respectively). Diagnostic confidence at 100Sn kVp was significantly higher than at 70 kVp or 150Sn kV (ADMIRE 3, P < 0.05; ADMIRE 5, P < 0.01) at both the 1/10th and 1/20th dose levels. Images of the 2 patients scanned with 100Sn kVp at the 1/10th and 1/20th dose levels were of diagnostic quality. Our study suggests that chest CT for the detection of pulmonary nodules can be performed with third-generation dual-source CT producing high image quality, sensitivity, and diagnostic confidence at a very low effective radiation dose of 0.06 mSv when using a single-energy protocol at 100 kVp with spectral shaping and when using advanced iterative reconstruction techniques.

6 August 2011 in European Radiology
Verified
Journal Article
Closed access
PDF file
AWAnna Winklehner
Anna Winklehner
CKChristoph Karlo
Christoph Karlo
GPGilbert Puippe
Gilbert Puippe
BSBernhard Schmidt
Bernhard Schmidt
TFThomas Flohr
Thomas Flohr
Hatem AlkadhiHatem Alkadhi
Hatem Alkadhi

28 August 2015 in Arthritis & Rheumatology
Verified
Journal Article
Open access
PDF file
YSYossra A. Suliman
Yossra A. Suliman
RDRucsandra Dobrota
Rucsandra Dobrota
DHDörte Huscher
Dörte Huscher
TNThi D. L. Nguyen‐Kim
Thi D. L. Nguyen‐Kim
BMBritta Maurer
Britta Maurer
ODOliver Distler
Oliver Distler
Objective: Validated methods for the screening and early diagnosis of systemic sclerosis (SSc; scleroderma)–related interstitial lung disease (ILD) are needed. The aim of this study was to evaluate the performance of pulmonary function tests (PFTs) compared with that of high‐resolution computed tomography (HRCT) of the chest for the detection of SSc‐related ILD in clinical practice, and to identify predictors of lung involvement that is functionally occult but significant on HRCT.Methods: Prospectively enrolled patients with SSc were assessed according to the European League Against Rheumatism (EULAR)/EULAR Scleroderma Trial and Research standards. The assessment included PFTs and HRCT. The HRCT images were evaluated in a blinded manner by 2 experienced radiologists. The performance parameters of PFTs for the diagnosis of SSc‐related ILD were calculated. Predictors of significant ILD as determined by HRCT in patients with normal forced vital capacity (FVC) values were identified through logistic regression.Results: Among the 102 patients, 64 (63.0%) showed significant ILD on HRCT, while only 27 (26.0%) had an FVC <80% of predicted, and 54 (53.0%) had a decrease in the results of at least 1 PFT. Forty (62.5%) of 64 patients with significant ILD on HRCT had a normal FVC value, translating into a high false‐negative rate. Notably, 5 of 40 patients with a normal FVC value had severe, functionally occult lung fibrosis; in 2 of these patients, the results of all of the PFTs were within normal limits. Patients with normal FVC values despite evidence of fibrosis on HRCT more frequently had anti–Scl‐70 antibodies and diffuse SSc and less frequently had anticentromere antibodies (ACAs) compared with patients with both normal FVC values and normal HRCT results.Conclusion: The derived evidence‐based data reveal a high risk of missing significant SSc‐related ILD when relying solely on PFTs. More comprehensive screening algorithms for early detection are warranted. In particular, additional imaging investigations for the early detection of SSc‐related ILD should be considered in ACA‐negative patients with normal FVC values.

7 March 2018 in Acta Oncologica
Verified
Journal Article
Closed access
MPMatea Pavic
Matea Pavic
MBMarta Bogowicz
Marta Bogowicz
XWXaver Würms
Xaver Würms
SGStefan Glatz
Stefan Glatz
TFTobias Finazzi
Tobias Finazzi
STStephanie Tanadini-Lang
Stephanie Tanadini-Lang
Background: Radiomics is a promising methodology for quantitative analysis and description of radiological images using advanced mathematics and statistics. Tumor delineation, which is still often done manually, is an essential step in radiomics, however, inter-observer variability is a well-known uncertainty in radiation oncology. This study investigated the impact of inter-observer variability (IOV) in manual tumor delineation on the reliability of radiomic features (RF). Methods: Three different tumor types (head and neck squamous cell carcinoma (HNSCC), malignant pleural mesothelioma (MPM) and non-small cell lung cancer (NSCLC)) were included. For each site, eleven individual tumors were contoured on CT scans by three experienced radiation oncologists. Dice coefficients (DC) were calculated for quantification of delineation variability. RF were calculated with an in-house developed software implementation, which comprises 1404 features: shape (n = 18), histogram (n = 17), texture (n = 137) and wavelet (n = 1232). The IOV of RF was studied using the intraclass correlation coefficient (ICC). An ICC >0.8 indicates a good reproducibility. For the stable RF, an average linkage hierarchical clustering was performed to identify classes of uncorrelated features. Results: Median DC was high for NSCLC (0.86, range 0.57–0.90) and HNSCC (0.72, 0.21–0.89), whereas it was low for MPM (0.26, 0–0.9) indicating substantial IOV. Stability rate of RF correlated with DC and depended on tumor site, showing a high stability in NSCLC (90% of total parameters), acceptable stability in HNSCC (59% of total parameters) and low stability in MPM (36% of total parameters). Shape features showed the weakest stability across all tumor types. Hierarchical clustering revealed 14 groups of correlated and stable features for NSCLC and 6 groups for both HNSCC and MPM. Conclusion: Inter-observer delineation variability has a relevant influence on radiomics analysis and is strongly influenced by tumor type. This leads to a reduced number of suitable imaging features.