Deep learning model for differentiating nasal cavity masses based on nasal endoscopy images.

BACKGROUND: Nasal polyps and inverted papillomas often look similar. Clinically, it is difficult to distinguish the masses by endoscopic examination. Therefore, in this study, we aimed to develop a deep learning algorithm for computer-aided diagnosis of nasal endoscopic images, which may provide a more accurate clinical diagnosis before pathologic confirmation of the nasal masses. METHODS: By performing deep learning of nasal endoscope images, we evaluated our computer-aided diagnosis system's assessment ability for nasal polyps and inverted papilloma and the feasibility of their clinical application. We used curriculum learning pre-trained with patches of nasal endoscopic images and full-sized images. The proposed model's performance for classifying nasal polyps, inverted papilloma, and normal tissue was analyzed using five-fold cross-validation. RESULTS: The normal scores for our best-performing network were 0.9520 for recall, 0.7900 for precision, 0.8648 for F1-score, 0.97 for the area under the curve, and 0.8273 for accuracy. For nasal polyps, the best performance was 0.8162, 0.8496, 0.8409, 0.89, and 0.8273, respectively, for recall, precision, F1-score, area under the curve, and accuracy. Finally, for inverted papilloma, the best performance was obtained for recall, precision, F1-score, area under the curve, and accuracy values of 0.5172, 0.8125, 0.6122, 0.83, and 0.8273, respectively. CONCLUSION: Although there were some misclassifications, the results of gradient-weighted class activation mapping were generally consistent with the areas under the curve determined by otolaryngologists. These results suggest that the convolutional neural network is highly reliable in resolving lesion locations in nasal endoscopic images.

Results of a 10-year hearing screening using automated auditory brainstem response in newborns: The two-step AABR method.

OBJECTIVES: Approximately 1-6 in every 1000 children are born with hearing loss. An automated auditory brainstem response (AABR) test is essential for screening newborns for abnormal hearing. At the tertiary hospital, we have been using a two-step AABR protocol for newborn hearing assessment since 2005. This study aimed to report the 10-year hearing screening results of newborns at the institution, and prove the efficacy of the two-step AABR protocol. METHODS: From August 2005 to January 2015, 3059 newborns were screened through AABR testing using the MASTER ABaer system. The first screening test was performed after the first 24 h of life. If a newborn was referred, the test was performed within 1 month after discharge from the hospital. The results were regarded as pass when the point optimized variance ratio was >3.5, using a stimulus level of 35 dB HL. When newborns were referred for the second AABR, they received follow-up tests including tympanometry, ABR, auditory steady-state response, and otoacoustic emission within 3 months. RESULTS: A total of 3059 newborns underwent newborn hearing screening tests over a period of 10 years. One hundred and twenty (3.9%) newborns were referred with the initial AABR, and 104 (3.4%) were referred with a subsequent AABR. Of the newborns, 42 (1.37%) were confirmed to have a bilateral hearing impairment. CONCLUSIONS: It is known that the referral rate for the AABR test is 3-4%, as recommended by the Joint Committee on Infant Hearing. Our data showed a referral rate of 3.4%. The two-step AABR test has been useful for screening hearing loss in newborns at tertiary hospital.

Image-guided recording system for spatial and temporal mapping of neuronal activities in brain slice.

In this study, we introduce the novel image-guided recording system (IGRS) for efficient interpretation of neuronal activities in the brain slice. IGRS is designed to combine microelectrode array (MEA) and optical coherence tomography at the customized upright microscope. It allows to record multi-site neuronal signals and image of the volumetric brain anatomy in a single body configuration. For convenient interconnection between a brain image and neuronal signals, we developed the automatic mapping protocol that enables us to project acquired neuronal signals on a brain image. To evaluate the performance of IGRS, hippocampal signals of the brain slice were monitored, and corresponding with two-dimensional neuronal maps were successfully reconstructed. Our results indicated that IGRS and mapping protocol can provide the intuitive information regarding long-term and multi-sites neuronal signals. In particular, the temporal and spatial mapping capability of neuronal signals would be a very promising tool to observe and analyze the massive neuronal activity and connectivity in MEA-based electrophysiological studies.

Lamellar keratoplasty using position-guided surgical needle and M-mode optical coherence tomography.

Deep anterior lamellar keratoplasty (DALK) is an emerging surgical technique for the restoration of corneal clarity and vision acuity. The big-bubble technique in DALK surgery is the most essential procedure that includes the air injection through a thin syringe needle to separate the dysfunctional region of the cornea. Even though DALK is a well-known transplant method, it is still challenged to manipulate the needle inside the cornea under the surgical microscope, which varies its surgical yield. Here, we introduce the DALK protocol based on the position-guided needle and M-mode optical coherence tomography (OCT). Depth-resolved 26-gage needle was specially designed, fabricated by the stepwise transitional core fiber, and integrated with the swept source OCT system. Since our device is feasible to provide both the position information inside the cornea as well as air injection, it enables the accurate management of bubble formation during DALK. Our results show that real-time feedback of needle end position was intuitionally visualized and fast enough to adjust the location of the needle. Through our research, we realized that position-guided needle combined with M-mode OCT is a very efficient and promising surgical tool, which also to enhance the accuracy and stability of DALK.

Smartphone-Based Endoscope System for Advanced Point-of-Care Diagnostics: Feasibility Study.

BACKGROUND: Endoscopic technique is often applied for the diagnosis of diseases affecting internal organs and image-guidance of surgical procedures. Although the endoscope has become an indispensable tool in the clinic, its utility has been limited to medical offices or operating rooms because of the large size of its ancillary devices. In addition, the basic design and imaging capability of the system have remained relatively unchanged for decades. OBJECTIVE: The objective of this study was to develop a smartphone-based endoscope system capable of advanced endoscopic functionalities in a compact size and at an affordable cost and to demonstrate its feasibility of point-of-care through human subject imaging. METHODS: We developed and designed to set up a smartphone-based endoscope system, incorporating a portable light source, relay-lens, custom adapter, and homebuilt Android app. We attached three different types of existing rigid or flexible endoscopic probes to our system and captured the endoscopic images using the homebuilt app. Both smartphone-based endoscope system and commercialized clinical endoscope system were utilized to compare the imaging quality and performance. Connecting the head-mounted display (HMD) wirelessly, the smartphone-based endoscope system could superimpose an endoscopic image to real-world view. RESULTS: A total of 15 volunteers who were accepted into our study were captured using our smartphone-based endoscope system, as well as the commercialized clinical endoscope system. It was found that the imaging performance of our device had acceptable quality compared with that of the conventional endoscope system in the clinical setting. In addition, images captured from the HMD used in the smartphone-based endoscope system improved eye-hand coordination between the manipulating site and the smartphone screen, which in turn reduced spatial disorientation. CONCLUSIONS: The performance of our endoscope system was evaluated against a commercial system in routine otolaryngology examinations. We also demonstrated and evaluated the feasibility of conducting endoscopic procedures through a custom HMD.

FAST: Size-Selective, Clog-Free Isolation of Rare Cancer Cells from Whole Blood at a Liquid-Liquid Interface.

Circulating tumor cells (CTCs) have great potential to provide minimally invasive ways for the early detection of cancer metastasis and for the response monitoring of various cancer treatments. Despite the clinical importance and progress of CTC-based cancer diagnostics, most of the current methods of enriching CTCs are difficult to implement in general hospital settings due to complex and time-consuming protocols. Among existing technologies, size-based isolation methods provide antibody-independent, relatively simple, and high throughput protocols. However, the clogging issues and lower than desired recovery rates and purity are the key challenges. In this work, inspired by antifouling membranes with liquid-filled pores in nature, clog-free, highly sensitive (95.9 ± 3.1% recovery rate), selective (>2.5 log depletion of white blood cells), rapid (>3 mL/min), and label-free isolation of viable CTCs from whole blood without prior sample treatment is achieved using a stand-alone lab-on-a-disc system equipped with fluid-assisted separation technology (FAST). Numerical simulation and experiments show that this method provides uniform, clog-free, ultrafast cell enrichment with pressure drops much less than in conventional size-based filtration, at 1 kPa. We demonstrate the clinical utility of the point-of-care detection of CTCs with samples taken from 142 patients suffering from breast, stomach, or lung cancer.

One-photon and two-photon stimulation of neurons in a microfluidic culture system.

In this study, we demonstrate a novel platform for optical stimulation of neural circuits combined with a microfluidic culture method and microelectrode array measurements. Neuron-on-a-chip was designed and fabricated to isolate axons without a soma or dendrite. Thus, it is readily able to manipulate the neuronal alignment and to investigate the neuronal activity at the locations we want to observe. We adapted the optical stimulation technique to the arranged neurons to generate the neuronal signals in a non-invasive fashion. A blue light-emitting diode and a femtosecond laser with 780 nm center wavelength were used for neuronal activation and the corresponding neuronal signals were measured by MEAs at the same time. We found that one-photon light via caged glutamate provoked periodic spiking. In contrast, the femtosecond pulse irradiation generated repetitive firing at constant rates. Response times of one-photon and two-photon stimulation were around 200 ms and 50 ms, respectively. We also quantified neural responses, by varying optical parameters such as exposure time and irradiation power.

Distinct Membrane Disruption Pathways Are Induced by 40-Residue ß-Amyloid Peptides.

Cellular membrane disruption induced by ß-amyloid (Aß) peptides has been considered one of the major pathological mechanisms for Alzheimer disease. Mechanistic studies of the membrane disruption process at a high-resolution level, on the other hand, are hindered by the co-existence of multiple possible pathways, even in simplified model systems such as the phospholipid liposome. Therefore, separation of these pathways is crucial to achieve an in-depth understanding of the Aß-induced membrane disruption process. This study, which utilized a combination of multiple biophysical techniques, shows that the peptide-to-lipid (P:L) molar ratio is an important factor that regulates the selection of dominant membrane disruption pathways in the presence of 40-residue Aß peptides in liposomes. Three distinct pathways (fibrillation with membrane content leakage, vesicle fusion, and lipid uptake through a temporarily stable ionic channel) become dominant in model liposome systems under specific conditions. These individual systems are characterized by both the initial states of Aß peptides and the P:L molar ratio. Our results demonstrated the possibility to generate simplified Aß-membrane model systems with a homogeneous membrane disruption pathway, which will benefit high-resolution mechanistic studies in the future. Fundamentally, the possibility of pathway selection controlled by P:L suggests that the driving forces for Aß aggregation and Aß-membrane interactions may be similar at the molecular level.