Smartphone-based artificial intelligence using a transfer learning algorithm for the detection and diagnosis of middle ear diseases: A retrospective deep learning study.

Background: Middle ear diseases such as otitis media and middle ear effusion, for which diagnoses are often delayed or misdiagnosed, are among the most common issues faced by clinicians providing primary care for children and adolescents. Artificial intelligence (AI) has the potential to assist clinicians in the detection and diagnosis of middle ear diseases through imaging. Methods: Otoendoscopic images obtained by otolaryngologists from Taipei Veterans General Hospital in Taiwan between Jany 1, 2011 to Dec 31, 2019 were collected retrospectively and de-identified. The images were entered into convolutional neural network (CNN) training models after data pre-processing, augmentation and splitting. To differentiate sophisticated middle ear diseases, nine CNN-based models were constructed to recognize middle ear diseases. The best-performing models were chosen and ensembled in a small CNN for mobile device use. The pretrained model was converted into the smartphone-based program, and the utility was evaluated in terms of detecting and classifying ten middle ear diseases based on otoendoscopic images. A class activation map (CAM) was also used to identify key features for CNN classification. The performance of each classifier was determined by its accuracy, precision, recall, and F1-score. Findings: A total of 2820 clinical eardrum images were collected for model training. The programme achieved a high detection accuracy for binary outcomes (pass/refer) of otoendoscopic images and ten different disease categories, with an accuracy reaching 98.0% after model optimisation. Furthermore, the application presented a smooth recognition process and a user-friendly interface and demonstrated excellent performance, with an accuracy of 97.6%. A fifty-question questionnaire related to middle ear diseases was designed for practitioners with different levels of clinical experience. The AI-empowered mobile algorithm's detection accuracy was generally superior to that of general physicians, resident doctors, and otolaryngology specialists (36.0%, 80.0% and 90.0%, respectively). Our results show that the proposed method provides sufficient treatment recommendations that are comparable to those of specialists. Interpretation: We developed a deep learning model that can detect and classify middle ear diseases. The use of smartphone-based point-of-care diagnostic devices with AI-empowered automated classification can provide real-world smart medical solutions for the diagnosis of middle ear diseases and telemedicine. Funding: This study was supported by grants from the Ministry of Science and Technology (MOST110-2622-8-075-001, MOST110-2320-B-075-004-MY3, MOST-110-2634-F-A49 -005, MOST110-2745-B-075A-001A and MOST110-2221-E-075-005), Veterans General Hospitals and University System of Taiwan Joint Research Program (VGHUST111-G6-11-2, VGHUST111c-140), and Taipei Veterans General Hospital (V111E-002-3).

Audiological and otologic manifestations of glutaric aciduria type I.

BACKGROUND: Glutaric aciduria type 1 (GA-1) is a rare disease connected with speech delay and neurological deficits. However, the audiological and otologic profiles of GA-1 have not yet been fully characterized. To our knowledge, this is the largest study of comprehensive audiological and otologic evaluation in patients with GA-1 to date. METHODS: Thirteen patients diagnosed with GA-1 between January 1994 and December 2019 with audiological, radiological and genetic manifestations were retrospectively analyzed. Hearing tests were performed in all patients. MRI was performed for radiological evaluation. RESULTS: Hearing loss was found in 76.9% (10/13) of GA-1 patients, including slight hearing loss in 46.1% (6/13) of patients, mild hearing loss in 15.4% (2/13) of patients, and moderate hearing loss in 7.7% (1/13) of patients. Normal hearing thresholds were seen in 23% (3/13) of patients. Patients with intensive care unit (ICU) admission history showed significantly worse hearing than those without (29.17 ± 12.47 vs 13.56 ± 3.93 dB HL, 95% CI 2.92-24.70, p = 0.0176). One patient had moderate sensorineural hearing loss and a past history of acute encephalopathic crisis. No usual causative gene mutations associated with hearing loss were found in these patients. MRI showed a normal vestibulocochlear apparatus and cochlear nerve. One patient with extensive injury of the basal ganglia on MRI after acute encephalopathic crisis was found to have moderate sensorineural hearing loss. Two patients with disability scores above 5 were found to have mild to moderate hearing impairment. No obvious correlation between macrocephaly and hearing loss was found. CONCLUSION: A high prevalence of hearing impairment is found in GA-1 patients. Adequate audiological evaluation is essential for these patients, especially for those after encephalopathic crises or with ICU admission history.

Improving the Lifetime of CsPbBr3 Perovskite in Water Using Self-Healing and Transparent Elastic Polymer Matrix.

This study developed a simple and efficient strategy to stabilize inorganic halide perovskite CsPbX3 at high relative humidity by embedding it into the matrix with elastic and self-healing features. The polymer matrix has a naturally hydrophobic characteristic of n-butyl acrylate segment (n-BA) and cross-linkable and healable moiety from N-(hydroxymethyl) acrylamide segment (NMA). It was chosen due to the provisions of both a surrounding protective layer for inorganic perovskite and elastic, as well as healing ability to the whole organic-inorganic composite. This fabricated CsPbBr3/PBA-co-PNMA composite was demonstrated to stably persist against the suffering from hydrolysis of perovskites when exposed to a high moisture environment. The PL intensity of the composite after crosslinking was found to be relatively stable after 30 days of exposure to air. Upon water immersion, the PL intensity of composite only showed a decrease of 32% after the first 6 h, then remained stable for 6 h afterward. Furthermore, this fabricated composite was not only flexible and relatively transparent but also exhibited excellent self-healing capability in ambient conditions (T = 25°C), in which the self-healing efficiency after 24 h was above 40%. The tensile strength and stretching ability of 5 wt% perovskite content in the random copolymer were observed to be 3.8 MPa and 553.5% respectively. Overall, flexible and self-healing properties combining with high luminescence characteristics are very promising materials for next-generation soft optical devices.

An Elastic Interfacial Transistor Enabled by Superhydrophobicity.

Enabling mechanical responsiveness in field-effect transistors (FETs) offers new technological opportunity beyond the reach of existing platforms. Here a new force-sensing concept is proposed by controlling the wettability of a semiconductor surface, referring to the interfacial field-effect transistors (IFETs). An IFET made by superhydrophobic semiconductor nanowires (NWs) sandwiched between a layer of 2D electron gas (2DEG) and a conductive Cassie-Baxter (CB) sessile droplet is designed. Following the hydrostatic deformation of the CB droplet upon mechanical stress, an extremely small elastic modulus of 820 pascals vertical to the substrate plane, or ≈100 times softer than Ecoflex rubbers, enabling an excellent stress detection limit down to <10 pascals and a stress sensitivity of 36 kPa-1 is proposed. The IFET exhibits an on/off current ratio exceeding 3 × 104 , as the carrier density profile at the NW/2DEG interface is modulated by a partially penetrated electrostatic field. This study demonstrates a versatile platform that bridges multiple macroscopic interfacial phenomena with nanoelectronic responses.

Mitochondrial DNA released by trauma induces neutrophil extracellular traps.

Neutrophil extracellular traps (NETs) are critical for anti-bacterial activity of the innate immune system. We have previously shown that mitochondrial damage-associated molecular patterns (mtDAMPs), including mitochondrial DNA (mtDNA), are released into the circulation after injury. We therefore questioned whether mtDNA is involved in trauma-induced NET formation. Treatment of human polymorphoneutrophils (PMN) with mtDNA induced robust NET formation, though in contrast to phorbol myristate acetate (PMA) stimulation, no NADPH-oxidase involvement was required. Moreover, formation of mtDNA-induced NETs was completely blocked by TLR9 antagonist, ODN-TTAGGG. Knowing that infective outcomes of trauma in elderly people are more severe than in young people, we measured plasma mtDNA and NET formation in elderly and young trauma patients and control subjects. MtDNA levels were significantly higher in the plasma of elderly trauma patients than young patients, despite lower injury severity scores in the elderly group. NETs were not visible in circulating PMN isolated from either young or old control subjects. NETs were however, detected in PMN isolated from young trauma patients and to a lesser extent from elderly patients. Stimulation by PMA induced widespread NET formation in PMN from both young volunteers and young trauma patients. NET response to PMA was much less pronounced in both elderly volunteers' PMN and in trauma patients' PMN. We conclude that mtDNA is a potent inducer of NETs that activates PMN via TLR9 without NADPH-oxidase involvement. We suggest that decreased NET formation in the elderly regardless of higher mtDNA levels in their plasma may result from decreased levels of TLR9 and/or other molecules, such as neutrophil elastase and myeloperoxidase that are involved in NET generation. Further study of the links between circulating mtDNA and NET formation may elucidate the mechanisms of trauma-related organ failure as well as the greater susceptibility to secondary infection in elderly trauma patients.

Mitochondrial damage-associated molecular patterns from fractures suppress pulmonary immune responses via formyl peptide receptors 1 and 2.

BACKGROUND: No known biologic mechanisms link tissue injury with pneumonia (PNA). Neutrophils (PMNs) are innate immune cells that clear bacteria from the lung by migration toward chemoattractants and killing bacteria in neutrophil extracellular traps (NETs). We predicted that tissue injury would suppress PMN antimicrobial function in the lung. We have also shown that mitochondria-derived damage-associated molecular pattern molecules from the bone can alter PMN phenotype and so hypothesized that formyl peptides (FPs) from fractures predispose to PNA by suppressing PMN activity in the lung. METHODS: Animal studies involved the following. (1) Rats were divided into three groups (10 per condition) as follows: (a) saline injection in the thigh (b) Staphylococcus aureus (SA, 3 × 10) injected intratracheally, or (c) pseudofracture (PsFx; bone supernatant injected in the thigh) plus intratracheally injected SA. (2) Rats were divided into four groups as follows: (a) control, (b) pulmonary contusion (PC), (c) PsFx, and (d) PC + PsFx. Bronchoalveolar lavage was performed 16 hours later. Clinical studies involved the following. (3) Human bone supernatant was assayed for its FP-receptor (FPR) stimulation. (4) Trauma patients' PMN (n = 32; mean ± SE Injury Severity Score [ISS], 27 ± 10) were assayed for chemotaxis (CTX) or treated with Phorbol 12-myristate 13-acetate (PMA, Phorbol ester) and analyzed for NET formation. RESULTS: In the animal studies, (1) SA was rapidly cleared by the uninjured mice and PsFx markedly suppressed lung bacterial clearance (p < 0.01). (2a) PC induces PMN traffic to the lung, but PsFx decreases PC-induced PMN traffic (p < 0.01). (2b) SA increased bronchoalveolar lavage PMN, and PsFx decreased that influx (p < 0.01). In the clinical studies, (3) bone supernatant activates PMN both via FPR-1 and FPR-2. (4) Trauma decreases PMN CTX to multiple chemokines. Circulating PMNs show NETs spontaneously after trauma, but maximal NET formation is markedly attenuated. CONCLUSION: Fractures may decrease lung bacterial clearance because FP suppresses PMN CTX to other chemoattractants via FPR-1/2. Trauma activates NETosis but suppresses maximal NETosis. Fractures decrease lung bacterial clearance by multiple mechanisms. PNA after fractures may reflect damage-associated molecular pattern-mediated suppression of PMN antimicrobial function in the lung.