Comparison of Anomaly Detectors: Context Matters.

Deep generative models are challenging the classical methods in the field of anomaly detection nowadays. Every newly published method provides evidence of outperforming its predecessors, sometimes with contradictory results. The objective of this article is twofold: to compare anomaly detection methods of various paradigms with a focus on deep generative models and identification of sources of variability that can yield different results. The methods were compared on popular tabular and image datasets. We identified that the main sources of variability are the experimental conditions: 1) the type of dataset (tabular or image) and the nature of anomalies (statistical or semantic) and 2) strategy of selection of hyperparameters, especially the number of available anomalies in the validation set. Methods perform differently in different contexts, i.e., under a different combination of experimental conditions together with computational time. This explains the variability of the previous results and highlights the importance of careful specification of the context in the publication of a new method. All our code and results are available for download.

Comparison of 4 femoral tunnel drilling techniques in anterior cruciate ligament reconstruction.

PURPOSE: The purpose of this study was to determine which femoral tunnel drilling technique most closely reproduces the anatomic femoral footprint and has acceptable tunnel length and tunnel orientation. METHODS: We divided 20 cadaveric knees into 4 equal groups. Arthroscopically, the anatomic femoral footprint was marked with an awl as the tunnel starting point. In group 1 the femoral tunnel was drilled through a tibial tunnel. In groups 2 and 3 the femoral tunnel was drilled through the anteromedial arthroscopy portal, with a rigid drill and flexible drill, respectively. In group 4 the femoral tunnel was drilled with the outside-in technique over a pin positioned with an arthroscopic femoral guide. Measurements of the tunnel length, aperture, and placement were taken from 3-dimensional computed tomography scans. RESULTS: Tunnel length for groups 1, 2, 3, and 4 averaged 42.08 mm, 37.73 mm, 28.92 mm, and 31.96 mm (P = .039). The mean coronal angle of the tunnels as measured from the line tangent to the posterior femoral condyles was 63.30°, 61.22°, 51.77°, and 45.00° (P = .007), and the mean distance from the inferior articular surface to the edge of the tunnel was 5.60 mm, 4.36 mm, 2.42 mm, and -0.63 mm (P = .008) for groups 1, 2, 3, and 4, respectively. There was no statistical difference in footprint length, width, area, or distance from the posterior articular margin. CONCLUSION: Drilling by the transtibial technique produces the most vertical and longest tunnels. Independent drilling techniques produce the most anatomic tunnels but at the expense of tunnel length. CLINICAL RELEVANCE: When the orthopaedic surgeon is performing ACL reconstruction, it is critical to achieve anatomic placement of the graft, as well as maintain appropriate tunnel length.