Deep learning discrimination of rheumatoid arthritis from osteoarthritis on hand radiography.

PURPOSE: To develop a deep learning model to distinguish rheumatoid arthritis (RA) from osteoarthritis (OA) using hand radiographs and to evaluate the effects of changing pretraining and training parameters on model performance. MATERIALS AND METHODS: A convolutional neural network was retrospectively trained on 9714 hand radiograph exams from 8387 patients obtained from 2017 to 2021 at seven hospitals within an integrated healthcare network. Performance was assessed using an independent test set of 250 exams from 146 patients. Binary discriminatory capacity (no arthritis versus arthritis; RA versus not RA) and three-way classification (no arthritis versus OA versus RA) were evaluated. The effects of additional pretraining using musculoskeletal radiographs, using all views as opposed to only the posteroanterior view, and varying image resolution on model performance were also investigated. Area under the receiver operating characteristic curve (AUC) and Cohen's kappa coefficient were used to evaluate diagnostic performance. RESULTS: For no arthritis versus arthritis, the model achieved an AUC of 0.975 (95% CI: 0.957, 0.989). For RA versus not RA, the model achieved an AUC of 0.955 (95% CI: 0.919, 0.983). For three-way classification, the model achieved a kappa of 0.806 (95% CI: 0.742, 0.866) and accuracy of 87.2% (95% CI: 83.2%, 91.2%) on the test set. Increasing image resolution increased performance up to 1024 × 1024 pixels. Additional pretraining on musculoskeletal radiographs and using all views did not significantly affect performance. CONCLUSION: A deep learning model can be used to distinguish no arthritis, OA, and RA on hand radiographs with high performance.

A contemporary analysis of Fournier gangrene using the National Surgical Quality Improvement Program.

OBJECTIVE: To determine a nationwide contemporary description of surgical Fournier gangrene (FG) and necrotizing fasciitis of the genitalia (NFG) outcomes because historically reported mortality rates for FG and NFG are based on small single-institution studies from the 1980s and the 1990s. METHODS: The National Surgical Quality Improvement Program is a risk-adjusted surgical database used by nearly 400 hospitals nationwide, which tracks preoperative, intraoperative, and 30-day postoperative clinical variables. Data are extracted from patient charts by an independent surgical clinical reviewer at each hospital. Using the National Surgical Quality Improvement Program data from 2005 to 2009, we calculated 30-day mortality rates and identified preoperative factors associated with increased mortality. RESULTS: A total of 650 patients were identified with surgery for FG or NFG. Fourteen patients with do not resuscitate orders placed preoperatively were excluded from analyses. For the remaining 636 patients, the overall 30-day mortality was 10.1% (64 of 636). Fifty-seven percent of patients (360 of 636) were men, 70% (446 of 636) were white, and 13% (81 of 636) were African American. Multivariate logistic regression indicated that increased age (odds ratio [OR], 1.041; P = .004), body mass index (OR, 1.045; P <.001), and preoperative white blood cell count (OR, 1.061; P = .001), and decreased platelet count (OR, 0.993; P <.001) were all associated with increased risk of death. CONCLUSION: We determined a surgical mortality rate for FG-NFG of 10.1%. This rate is about half of historically published estimates and similar to recent studies. The lower rate may indicate improvements in therapy. Increased age, body mass index, and white blood cell count, and decreased platelet count were all associated with an increased risk of 30-day mortality.

Epicutaneous model of community-acquired Staphylococcus aureus skin infections.

Staphylococcus aureus is one of the most common etiological agents of community-acquired skin and soft tissue infection (SSTI). Although the majority of S. aureus community-acquired SSTIs are uncomplicated and self-clearing in nature, some percentage of these cases progress into life-threatening invasive infections. Current animal models of S. aureus SSTI suffer from two drawbacks: these models are a better representation of hospital-acquired SSTI than community-acquired SSTI, and they involve methods that are difficult to replicate. For these reasons, we sought to develop a murine model of community-acquired methicillin-resistant S. aureus SSTI (CA-MRSA SSTI) that can be consistently reproduced with a high degree of precision. We utilized this model to begin to characterize the host immune response to this type of infection. We infected mice via epicutaneous challenge of the skin on the outer ear pinna using Morrow-Brown allergy test needles coated in S. aureus USA300. When mice were challenged in this model, they developed small, purulent, self-clearing lesions with predictable areas of inflammation that mimicked a human infection. CFU in the ear pinna peaked at day 7 before dropping by day 14. The T(h)1 and T(h)17 cytokines gamma interferon (IFN-γ), interleukin-12 (IL-12) p70, tumor necrosis factor alpha (TNF-α), IL-17A, IL-6, and IL-21 were all significantly increased in the draining lymph node of infected mice, and there was neutrophil recruitment to the infection site. In vivo neutrophil depletion demonstrated that neutrophils play a protective role in preventing bacterial dissemination and fatal invasive infection.

Updating of the spatial reference frame of head direction cells in response to locomotion in the vertical plane.

Many species navigate in three dimensions and are required to maintain accurate orientation while moving in an Earth vertical plane. Here we explored how head direction (HD) cells in the rat anterodorsal thalamus responded when rats locomoted along a 360° spiral track that was positioned vertically within the room at the N, S, E, or W location. Animals were introduced into the vertical plane either through passive placement (experiment 1) or by allowing them to run up a 45° ramp from the floor to the vertically positioned platform (experiment 2). In both experiments HD cells maintained direction-specific firing in the vertical plane with firing properties that were indistinguishable from those recorded in the horizontal plane. Interestingly, however, the cells' preferred directions were linked to different aspects of the animal's environment and depended on how the animal transitioned into the vertical plane. When animals were passively placed onto the vertical surface, the cells switched from using the room (global cues) as a reference frame to using the vertically positioned platform (local cues) as a reference frame, independent of where the platform was located. In contrast, when animals self-locomoted into the vertical plane, the cells' preferred directions remained anchored to the three-dimensional room coordinates and their activity could be accounted for by a simple 90° rotation of the floor's horizontal coordinate system to the vertical plane. These findings highlight the important role that active movement signals play for maintaining and updating spatial orientation when moving in three dimensions.