ABSTRACT
Many people with some form of hearing loss consider lipreading as their primary mode of day-to-day communication. However, finding resources to learn or improve one's lipreading skills can be challenging. This is further exacerbated in the COVID19 pandemic due to restrictions on direct interactions with peers and speech therapists. Today, online MOOCs platforms like Coursera and Udemy have become the most effective form of training for many types of skill development. However, online lipreading resources are scarce as creating such resources is an extensive process needing months of manual effort to record hired ac-tors. Because of the manual pipeline, such platforms are also limited in vocabulary, supported languages, accents, and speakers and have a high usage cost. In this work, we investigate the possibility of replacing real human talking videos with synthetically generated videos. Synthetic data can easily incorporate larger vocabularies, variations in accent, and even local languages and many speakers. We propose an end-to-end automated pipeline to develop such a platform using state-of-the-art talking head video generator networks, text-to-speech models, and computer vision techniques. We then perform an extensive human evaluation using carefully thought out lipreading exercises to validate the quality of our designed platform against the existing lipreading platforms. Our studies concretely point toward the potential of our approach in developing a large-scale lipreading MOOC platform that can impact millions of people with hearing loss. © 2023 IEEE.
ABSTRACT
Background. Corynebacterium striatum (CS), a common human commensal colonizing the skin and nasopharynx, has been associated with nosocomial infections in immunocompromised and chronically ill patients. During the winter 2020-2021 COVID-19 surge, a 420-bed California hospital reported a marked increase in CS respiratory cultures among ventilated COVID-19 patients. We conducted a public health investigation to assess and mitigate nosocomial transmission and contributing infection prevention and control (IPC) practices. Methods. A case was defined as a patient with CS in respiratory cultures from January 1, 2020 - February 28, 2021. We reviewed clinical characteristics on a subset of cases in 2021 and IPC practices in affected hospital locations. CS respiratory isolates collected on different dates and locations were assessed for relatedness by whole genome sequencing (WGS) on MiSeq. Results. Eighty-three cases were identified, including 75 among COVID-19 patients (Figure 1). Among 62 patients identified in 2021, all were ventilated;58 also had COVID-19, including 4 cases identified on point prevalence survey (PPS). The median time from admission to CS culture was 19 days (range, 0-60). Patients were critically ill;often it was unclear whether CS cultures represented colonization or infection. During the COVID-19 surge, two hospital wings (7W and 7S) were converted to negative-pressure COVID-19 units. Staff donned and doffed personal protective equipment in anterooms outside the units;extended use of gowns was practiced, and lapses in glove changes and hand hygiene (HH) between patients likely occurred. In response to the CS outbreak, patients were placed in Contact precautions and cohorted. Staff were re-educated on IPC for COVID-19 patients. Gowns were changed between CS patients. Subsequent PPS were negative. Two CS clusters were identified by WGS: cluster 1 (5 cases) in unit 7W, and cluster 2 (2 cases) in unit 7S (Figure 2). Conclusion. A surge in patients, extended use of gowns and lapses in core IPC practices including HH and environmental cleaning and disinfection during the winter 2020-2021 COVID-19 surge likely contributed to this CS outbreak. WGS provides supportive evidence for nosocomial CS transmission among critically ill COVID-19 patients.