Semi-automated De-identification of German Content Sensitive Reports for Big Data Analytics.

Purpose Projects involving collaborations between different institutions require data security via selective de-identification of words or phrases. A semi-automated de-identification tool was developed and evaluated on different types of medical reports natively and after adapting the algorithm to the text structure. Materials and Methods A semi-automated de-identification tool was developed and evaluated for its sensitivity and specificity in detecting sensitive content in written reports. Data from 4671 pathology reports (4105 + 566 in two different formats), 2804 medical reports, 1008 operation reports, and 6223 radiology reports of 1167 patients suffering from breast cancer were de-identified. The content was itemized into four categories: direct identifiers (name, address), indirect identifiers (date of birth/operation, medical ID, etc.), medical terms, and filler words. The software was tested natively (without training) in order to establish a baseline. The reports were manually edited and the model re-trained for the next test set. After manually editing 25, 50, 100, 250, 500 and if applicable 1000 reports of each type re-training was applied. Results In the native test, 61.3 % of direct and 80.8 % of the indirect identifiers were detected. The performance (P) increased to 91.4 % (P25), 96.7 % (P50), 99.5 % (P100), 99.6 % (P250), 99.7 % (P500) and 100 % (P1000) for direct identifiers and to 93.2 % (P25), 97.9 % (P50), 97.2 % (P100), 98.9 % (P250), 99.0 % (P500) and 99.3 % (P1000) for indirect identifiers. Without training, 5.3 % of medical terms were falsely flagged as critical data. The performance increased, after training, to 4.0 % (P25), 3.6 % (P50), 4.0 % (P100), 3.7 % (P250), 4.3 % (P500), and 3.1 % (P1000). Roughly 0.1 % of filler words were falsely flagged. Conclusion Training of the developed de-identification tool continuously improved its performance. Training with roughly 100 edited reports enables reliable detection and labeling of sensitive data in different types of medical reports. Key Points: · Collaborations between different institutions require de-identification of patients' data. · Software-based de-identification of content-sensitive reports grows in importance as a result of 'Big data'. · A de-identification software was developed and tested natively and after training. · The proposed de-identification software worked quite reliably, following training with roughly 100 edited reports. · A final check of the texts by an authorized person remains necessary. Citation Format · Seuss H, Dankerl P, Ihle M et al. Semi-automated De-identification of German Content Sensitive Reports for Big Data Analytics. Fortschr Röntgenstr 2017; 189: 661 - 671.

Deltoid Tuberosity Index: A Simple Radiographic Tool to Assess Local Bone Quality in Proximal Humerus Fractures.

BACKGROUND: Osteoporosis may complicate surgical fixation and healing of proximal humerus fractures and should be assessed preoperatively. Peripheral quantitative CT (pQCT) and the Tingart measurement are helpful methods, but both have limitations in clinical use because of limited availability (pQCT) or fracture lines crossing the area of interest (Tingart measurement). The aim of our study was to introduce and validate a simple cortical index to assess the quality of bone in proximal humerus fractures using AP radiographs. QUESTIONS/PURPOSES: We asked: (1) How do the deltoid tuberosity index and Tingart measurement correlate with each other, with patient age, and local bone mineral density (BMD) of the humeral head, measured by pQCT? (2) Which threshold values for the deltoid tuberosity index and Tingart measurement optimally discriminate poor local bone quality of the proximal humerus? (3) Are the deltoid tuberosity index and Tingart measurement clinically applicable and reproducible in patients with proximal humerus fractures? METHODS: The deltoid tuberosity index was measured immediately above the upper end of the deltoid tuberosity. At this position, where the outer cortical borders become parallel, the deltoid tuberosity index equals the ratio between the outer cortical and inner endosteal diameter. In the first part of our study, we retrospectively measured the deltoid tuberosity index on 31 patients (16 women, 15 men; mean age, 65 years; range, 22-83 years) who were scheduled for elective surgery other than fracture repair. Inclusion criteria were available native pQCT scans, AP shoulder radiographs taken in internal rotation, and no previous shoulder surgery. The deltoid tuberosity index and the Tingart measurement were measured on the preoperative internal rotation AP radiograph. The second part of our study was performed by reviewing 40 radiographs of patients with proximal humerus fractures (31 women, nine men; median age, 65 years; range, 22-88 years). Interrater (two surgeons) and intrarater (two readings) reliabilities, applicability, and diagnostic accuracy were assessed. RESULTS: The correlations between radiograph measurements and local BMD (pQCT) were strong for the deltoid tuberosity index (r = 0.80; 95% CI, 0.63-0.90; p < 0.001) and moderate for the Tingart measurement (r = 0.67; 95% CI, 0.42-0.83; p < 0.001). There was moderate correlation between patient age and the deltoid tuberosity index (r = 0.65; p < 0.001), patient age and the Tingart measurement (r = 0.69; p < 0.001), and patient age and pQCT (r = 0.73; p < 0.001). The correlation between the deltoid tuberosity index and the Tingart measurement was strong (r = 0.84; p < 0.001). We determined the cutoff value for the deltoid tuberosity index to be 1.44, with the area under the curve = 0.87 (95% CI, 0.74-0.99). This provided a sensitivity of 0.88 and specificity of 0.80. For the Tingart measurement, we determined the cutoff value to be 5.3 mm, with the area under the curve = 0.83 (95% CI, 0.67-0.98), which resulted in a sensitivity of 0.81 and specificity of 0.85. The intraobserver reliability was high and not different between the Tingart measurement (intraclass correlation coefficients [ICC] = 0.75 and 0.88) and deltoid tuberosity index (ICC = 0.88 and 0.82). However, interobserver reliability was higher for the deltoid tuberosity index (ICC = 0.96; 95% CI, 0.93-0.98) than for the Tingart measurement (ICC = 0.85; 95% CI, 0.69-0.93).The clinical applicability on AP radiographs of fractures was better for the deltoid tuberosity index (p = 0.025) because it was measureable on more of the radiographs (77/80; 96%) than the Tingart measurement (69/80; 86%). CONCLUSIONS: The deltoid tuberosity index correlated strongly with local BMD measured on pQCT and our study evidence shows that it is a reliable, simple, and applicable tool to assess local bone quality in the proximal humerus. We found that deltoid tuberosity index values consistently lower than 1.4 indicated low local BMD of the proximal humerus. Furthermore, the use of the deltoid tuberosity index has important advantages over the Tingart measurement regarding clinical applicability in patients with proximal humerus fractures, when fracture lines obscure the Tingart measurement landmarks. However, further studies are needed to assess the effect of the deltoid tuberosity index measurement and osteoporosis on treatment and outcome in patients with proximal humerus fractures. LEVEL OF EVIDENCE: Level IV, diagnostic study.