Generalizability of Deep Learning Classification of Spinal Osteoporotic Compression Fractures on Radiographs Using an Adaptation of the Modified-2 Algorithm-Based Qualitative Criteria.

RATIONALE AND OBJECTIVES: Spinal osteoporotic compression fractures (OCFs) can be an early biomarker for osteoporosis but are often subtle, incidental, and underreported. To ensure early diagnosis and treatment of osteoporosis, we aimed to build a deep learning vertebral body classifier for OCFs as a critical component of our future automated opportunistic screening tool. MATERIALS AND METHODS: We retrospectively assembled a local dataset, including 1790 subjects and 15,050 vertebral bodies (thoracic and lumbar). Each vertebral body was annotated using an adaption of the modified-2 algorithm-based qualitative criteria. The Osteoporotic Fractures in Men (MrOS) Study dataset provided thoracic and lumbar spine radiographs of 5994 men from six clinical centers. Using both datasets, five deep learning algorithms were trained to classify each individual vertebral body of the spine radiographs. Classification performance was compared for these models using multiple metrics, including the area under the receiver operating characteristic curve (AUC-ROC), sensitivity, specificity, and positive predictive value (PPV). RESULTS: Our best model, built with ensemble averaging, achieved an AUC-ROC of 0.948 and 0.936 on the local dataset's test set and the MrOS dataset's test set, respectively. After setting the cutoff threshold to prioritize PPV, this model achieved a sensitivity of 54.5% and 47.8%, a specificity of 99.7% and 99.6%, and a PPV of 89.8% and 94.8%. CONCLUSION: Our model achieved an AUC-ROC>0.90 on both datasets. This testing shows some generalizability to real-world clinical datasets and a suitable performance for a future opportunistic osteoporosis screening tool.

Effective Modulation of Male Aggression through Lateral Septum to Medial Hypothalamus Projection.

Aggression is a prevalent behavior in the animal kingdom that is used to settle competition for limited resources. Given the high risk associated with fighting, the central nervous system has evolved an active mechanism to modulate its expression. Lesioning the lateral septum (LS) is known to cause "septal rage," a phenotype characterized by a dramatic increase in the frequency of attacks. To understand the circuit mechanism of LS-mediated modulation of aggression, we examined the influence of LS input on the cells in and around the ventrolateral part of the ventromedial hypothalamus (VMHvl)-a region required for male mouse aggression. We found that the inputs from the LS inhibited the attack-excited cells but surprisingly increased the overall activity of attack-inhibited cells. Furthermore, optogenetic activation of the projection from LS cells to the VMHvl terminated ongoing attacks immediately but had little effect on mounting. Thus, LS projection to the ventromedial hypothalamic area represents an effective pathway for suppressing male aggression.

Proteomic characterization of Yersinia pestis virulence.

The Yersinia pestis proteome was studied as a function of temperature and calcium by two-dimensional differential gel electrophoresis. Over 4,100 individual protein spots were detected, of which hundreds were differentially expressed. A total of 43 differentially expressed protein spots, representing 24 unique proteins, were identified by mass spectrometry. Differences in expression were observed for several virulence-associated factors, including catalase-peroxidase (KatY), murine toxin (Ymt), plasminogen activator (Pla), and F1 capsule antigen (Caf1), as well as several putative virulence factors and membrane-bound and metabolic proteins. Differentially expressed proteins not previously reported to contribute to virulence are candidates for more detailed mechanistic studies, representing potential new virulence determinants.

Proteomic analysis of human serum by two-dimensional differential gel electrophoresis after depletion of high-abundant proteins.

Two-dimensional differential gel electrophoresis (2-D DIGE) was used to analyze human serum following the removal of albumin and five other high-abundant serum proteins. After protein removal, serum was analyzed by SDS-PAGE as a preliminary screen, and significant differences between four high-abundant protein removal methods were observed. Antibody-based albumin removal and high-abundant protein removal methods were found to be efficient and specific. To further characterize serum after protein removal, 2-D DIGE was employed, enabling multiplexed analysis of serum through the use of three fluorescent protein dyes. Comparison between crude serum and serum after removal of high-abundant proteins clearly illustrates an increase in the number of lower abundant protein spots observed. Approximately 850 protein spots were detected in crude serum whereas over 1500 protein spots were exposed following removal of six high-abundant proteins, representing a 76% increase in protein spot detection. Several proteins that showed a 2-fold increase in intensity after depletion of high-abundant proteins, as well as proteins that were depleted during abundant protein removal methods, were further characterized by mass spectrometry. This series of experiments demonstrates that high-abundant protein removal, combined with 2-D DIGE, is a practical approach for enriching and characterizing lower abundant proteins in human serum. Consequently, this methodology offers advances in proteomic characterization, and therefore, in the identification of biomarkers from human serum.