Reduced fetal cerebral blood flow predicts perinatal mortality in a mouse model of prenatal alcohol and cannabinoid exposure.

BACKGROUND: Children exposed prenatally to alcohol or cannabinoids individually can exhibit growth deficits and increased risk for adverse birth outcomes. However, these drugs are often co-consumed and their combined effects on early brain development are virtually unknown. The blood vessels of the fetal brain emerge and mature during the neurogenic period to support nutritional needs of the rapidly growing brain, and teratogenic exposure during this gestational window may therefore impair fetal cerebrovascular development. STUDY DESIGN: To determine whether prenatal polysubstance exposure confers additional risk for impaired fetal-directed blood flow, we performed high resolution in vivo ultrasound imaging in C57Bl/6J pregnant mice. After pregnancy confirmation, dams were randomly assigned to one of four groups: drug-free control, alcohol-exposed, cannabinoid-exposed or alcohol-and-cannabinoid-exposed. Drug exposure occurred daily between Gestational Days 12-15, equivalent to the transition between the first and second trimesters in humans. Dams first received an intraperitoneal injection of either cannabinoid agonist CP-55,940 (750 µg/kg) or volume-equivalent vehicle. Then, dams were placed in vapor chambers for 30 min of inhalation of either ethanol or room air. Dams underwent ultrasound imaging on three days of pregnancy: Gestational Day 11 (pre-exposure), Gestational Day 13.5 (peri-exposure) and Gestational Day 16 (post-exposure). RESULTS: All drug exposures decreased fetal cranial blood flow 24-hours after the final exposure episode, though combined alcohol and cannabinoid co-exposure reduced internal carotid artery blood flow relative to all other exposures. Umbilical artery metrics were not affected by drug exposure, indicating a specific vulnerability of fetal cranial circulation. Cannabinoid exposure significantly reduced cerebroplacental ratios, mirroring prior findings in cannabis-exposed human fetuses. Post-exposure cerebroplacental ratios significantly predicted subsequent perinatal mortality (p = 0.019, area under the curve, 0.772; sensitivity, 81%; specificity, 85.70%) and retroactively diagnosed prior drug exposure (p = 0.005; AUC, 0.861; sensitivity, 86.40%; specificity, 66.7%). CONCLUSIONS: Fetal cerebrovasculature is significantly impaired by exposure to alcohol or cannabinoids, and co-exposure confers additional risk for adverse birth outcomes. Considering the rising potency and global availability of cannabis products, there is an imperative for research to explore translational models of prenatal drug exposure, including polysubstance models, to inform appropriate strategies for treatment and care in pregnancies affected by drug exposure.

Mel frequency spectral domain defenses against adversarial attacks on speech recognition systems.

Automatic speech recognition (ASR) systems are vulnerable to adversarial attacks due to their reliance on machine learning models. Many of the defenses explored for defending ASR systems simply adapt defense approaches developed for the image domain. This paper explores speech-specific defenses in the feature domain and introduces a defense method called mel domain noise flooding (MDNF). MDNF injects additive noise to the mel spectrogram speech representation prior to re-synthesizing the audio signal input to ASR. The defense is evaluated against strong white-box threat models and shows competitive robustness.