Comparison of Pre-Trained YOLO Models on Steel Surface Defects Detector Based on Transfer Learning with GPU-Based Embedded Devices.

Steel is one of the most basic ingredients, which plays an important role in the machinery industry. However, the steel surface defects heavily affect its quality. The demand for surface defect detectors draws much attention from researchers all over the world. However, there are still some drawbacks, e.g., the dataset is limited accessible or small-scale public, and related works focus on developing models but do not deeply take into account real-time applications. In this paper, we investigate the feasibility of applying stage-of-the-art deep learning methods based on YOLO models as real-time steel surface defect detectors. Particularly, we compare the performance of YOLOv5, YOLOX, and YOLOv7 while training them with a small-scale open-source NEU-DET dataset on GPU RTX 2080. From the experiment results, YOLOX-s achieves the best accuracy of 89.6% mAP on the NEU-DET dataset. Then, we deploy the weights of trained YOLO models on Nvidia devices to evaluate their real-time performance. Our experiments devices consist of Nvidia Jetson Nano and Jetson Xavier AGX. We also apply some real-time optimization techniques (i.e., exporting to TensorRT, lowering the precision to FP16 or INT8 and reducing the input image size to 320 × 320) to reduce detection speed (fps), thus also reducing the mAP accuracy.

A deep learning algorithm may automate intracranial aneurysm detection on MR angiography with high diagnostic performance.

OBJECTIVES: To develop a deep learning algorithm for automated detection and localization of intracranial aneurysms on time-of-flight MR angiography and evaluate its diagnostic performance. METHODS: In a retrospective and multicenter study, MR images with aneurysms based on radiological reports were extracted. The examinations were randomly divided into two data sets: training set of 468 examinations and internal test set of 120 examinations. Additionally, 50 examinations without aneurysms were randomly selected and added to the internal test set. External test data set consisted of 56 examinations with intracranial aneurysms and 50 examinations without aneurysms, which were extracted based on radiological reports from a different institution. After manual ground truth segmentation of aneurysms, a deep learning algorithm based on 3D ResNet architecture was established with the training set. Its sensitivity, positive predictive value, and specificity were evaluated in the internal and external test sets. RESULTS: MR images included 551 aneurysms (mean diameter, 4.17 ± 2.49 mm) in the training, 147 aneurysms (mean diameter, 3.98 ± 2.11 mm) in the internal test, 63 aneurysms (mean diameter, 3.23 ± 1.69 mm) in the external test sets. The sensitivity, the positive predictive value, and the specificity were 87.1%, 92.8%, and 92.0% for the internal test set and 85.7%, 91.5%, and 98.0% for the external test set, respectively. CONCLUSION: A deep learning algorithm detected intracranial aneurysms with a high diagnostic performance which was validated using external data set. KEY POINTS: • A deep learning-based algorithm for the automated diagnosis of intracranial aneurysms demonstrated a high sensitivity, positive predictive value, and specificity. • The high diagnostic performance of the algorithm was validated using external test data set from a different institution with a different scanner. • The algorithm might be robust and effective for general use in real clinical settings.

Clinical results of the open ring PMMA guider assisted capsulorrhexis in cataract surgery.

BACKGROUND: To compare the results of continuous curvilinear capsulorrhexis(CCC) after application of an open ring-shaped guider compared with a free-hand procedure in eyes with cataracts. METHODS: This study comprised patients undergoing cataract surgery in Seoul St.Mary's Hospital, The Catholic University of Korea. Eyes were grouped depending on the capsulotomy method; CCC was performed by free-hand procedure on 94 eyes (free-hand group), and it was performed under the guidance after introduction of an open ring-shaped guider on consecutive 89 eyes (guided group). Horizontal and vertical diameter, area and circularity of capsulotomy were measured postoperatively at one day, two months and six months. Differences in parameters and the percentage of ideal capsulorrhexis were analyzed between the two groups. RESULTS: On the first postoperative day, the vertical diameter in the guided group (5.24 ± 0.16 mm) was significantly longer than that of the free-hand group (5.01 ± 0.65 mm, P = 0.019). The area of capsulotomy was larger in the guided group (21.55 ± 0.87 mm2) than that of the free-hand group (20.34 ± 2.96 mm2, P < 0.001). Circularity in the guided group (0.84 ± 0.03), was significantly greater than that of the free-hand group (0.69 ± 0.17, P = 0.036). Ideal capsulorrhexis was obtained in 60 eyes (67%) in the free-hand group and 81 eyes (86%) in the guided group. CONCLUSIONS: After introduction of an open ring-shaped guider, CCC became larger and more circular with less anterior capsular contracture. The rate of acquiring ideal capsulorrhexis was higher in the guided group than it was in the free-hand group for six months after surgery.

Ideal parameters for femto-second laser-assisted anterior capsulotomy: Animal studies.

In femtosecond laser-assisted cataract surgery, the parameter such as horizontal spot spacing and energy level can be adjusted. Although there have been several studies reported on various laser systems, showing the effects of varying energy levels and horizontal spot spacing on lens capsulotomy cut edges, none have been reported on the Catalys laser system (Abbott Medical Optics, Inc., Santa Ana, CA). The aim of this study is to evaluate, using scanning electron microscopy (SEM), the quality of the cut edges of the laser lens capsulotomy obtained using the Catalys Laser System, using different horizontal spot spacing and energy levels, and to determine the ideal parameters based on SEM results. Fifty rabbit capsulorhexis specimens from a femtosecond laser with different spot spacing and energy settings were divided into five groups randomly. Spot spacing was 3 um and laser pulse energy was 4 uJ in group 1. The respective values were 5 um and 2 uJ in group 2, 5 um and 4 uJ in group 3, 5 um and 6 uJ in group 4, and 7 um and 4 uJ in group 5. All samples were evaluated using SEM to compare the number of tags per capsulotomy and the laser emission time. Group 1 had a significantly lower tag formation than groups 3 and 5 (P = 0.042 and 0.021, respectively). Although the laser emission time increased about 1.5 sec as the spot spacing increased from 3 to 7 um, the quality of the cut was smoother in group 1 because of overlapping effect of photodisruption cavities. There was no significant difference between groups 2, 3 and 4 at different laser energy settings. In an ex-vivo study, samples from an energy setting of 10 uJ showed increased irregularity and damage. The degree of irregularity was higher at increasing spot spacing and laser energy settings, with abundant tag formation. Dense spot spacing with low-energy settings provide a better cut quality, which is probably correlated with the reduction in anterior capsular tear complications.

Highly Sensitive Ammonia Gas Sensor Based on Single-Crystal Poly(3-hexylthiophene) (P3HT) Organic Field Effect Transistor.

A highly sensitive organic field-effect transistor (OFET)-based sensor for ammonia in the range of 0.01 to 25 ppm was developed. The sensor was fabricated by employing an array of single-crystal poly(3-hexylthiophene) (P3HT) nanowires as the organic semiconductor (OSC) layer of an OFET with a top-contact geometry. The electrical characteristics (field-effect mobility, on/off current ratio) of the single-crystal P3HT nanowire OFET were about 2 orders of magnitude larger than those of the P3HT thin film OFET with the same geometry. The P3HT nanowire OFET showed excellent sensitivity to ammonia, about 3 times higher than that of the P3HT thin film OFET at 25 ppm ammonia. The ammonia response of the OFET was reversible and was not affected by changes in relative humidity from 45 to 100%. The high ammonia sensitivity of the P3HT nanowire OFET is believed to result from the single crystal nature and high surface/volume ratio of the P3HT nanowire used in the OSC layer.

Matrix Density Engineering of Hydrogel Nanoparticles with Simulation-Guided Synthesis for Tuning Drug Release and Cellular Uptake.

The use of a nanoparticle (NP)-based antitumor drug carrier has been an emerging strategy for selectively delivering the drugs to the tumor area and, thus, reducing the side effects that are associated with a high systemic dose of antitumor drugs. Precise control of drug loading and release is critical so as to maximize the therapeutic index of the NPs. Here, we propose a simple method of synthesizing NPs with tunable drug release while maintaining their loading ability, by varying the polymer matrix density of amine- or carboxyl-functionalized hydrogel NPs. We find that the NPs with a loose matrix released more cisplatin, with up to a 33 times faster rate. Also, carboxyl-functionalized NPs loaded more cisplatin and released it at a faster rate than amine-functionalized NPs. We performed detailed Monte Carlo computer simulations that elucidate the relation between the matrix density and drug release kinetics. We found good agreement between the simulation model and the experimental results for drug release as a function of time. Also, we compared the cellular uptake between amine-functionalized NPs and carboxyl-functionalized NPs, as a higher cellular uptake of NPs leads to improved cisplatin delivery. The amine-functionalized NPs can deliver 3.5 times more cisplatin into cells than the carboxyl-functionalized NPs. The cytotoxic efficacy of both the amine-functionalized NPs and the carboxyl-functionalized NPs showed a strong correlation with the cisplatin release profile, and the latter showed a strong correlation with the NP matrix density.

Extremely High Barrier Performance of Organic-Inorganic Nanolaminated Thin Films for Organic Light-Emitting Diodes.

This work presents a novel barrier thin film based on an organic-inorganic nanolaminate, which consists of alternating nanolayers of self-assembled organic layers (SAOLs) and Al2O3. The SAOLs-Al2O3 nanolaminated films were deposited using a combination of molecular layer deposition and atomic layer deposition techniques at 80 °C. Modulation of the relative thickness ratio of the SAOLs and Al2O3 enabled control over the elastic modulus and stress in the films. Furthermore, the SAOLs-Al2O3 thin film achieved a high degree of mechanical flexibility, excellent transmittance (>95%), and an ultralow water-vapor transmission rate (2.99 × 10-7 g m-2 day-1), which represents one of the lowest permeability levels ever achieved by thin film encapsulation. On the basis of its outstanding barrier properties with high flexibility and transparency, the nanolaminated film was applied to a commercial OLEDs panel as a gas-diffusion barrier film. The results showed defect propagation could be significantly inhibited by incorporating the SAOLs layers, which enhanced the durability of the panel.

Heterogeneous Monolithic Integration of Single-Crystal Organic Materials.

Manufacturing high-performance organic electronic circuits requires the effective heterogeneous integration of different nanoscale organic materials with uniform morphology and high crystallinity in a desired arrangement. In particular, the development of high-performance organic electronic and optoelectronic devices relies on high-quality single crystals that show optimal intrinsic charge-transport properties and electrical performance. Moreover, the heterogeneous integration of organic materials on a single substrate in a monolithic way is highly demanded for the production of fundamental organic electronic components as well as complex integrated circuits. Many of the various methods that have been designed to pattern multiple heterogeneous organic materials on a substrate and the heterogeneous integration of organic single crystals with their crystal growth are described here. Critical issues that have been encountered in the development of high-performance organic integrated electronics are also addressed.

Observation of Charge Separation and Space-Charge Region in Single-Crystal P3HT/C60 Heterojunction Nanowires.

We directly observed charge separation and a space-charge region in an organic single-crystal p-n heterojunction nanowire, by means of scanning photocurrent microscopy. The axial p-n heterojunction nanowire had a well-defined planar junction, consisted of P3HT (p-type) and C60 (n-type) single crystals and was fabricated by means of the recently developed inkjet-assisted nanotransfer printing technique. The depletion region formed at the p-n junction was directly observed by exploring the spatial distribution of photogenerated carriers along the heterojunction nanowire under various applied bias voltages. Our study provides a facile approach toward the precise characterization of charge transport in organic heterojunction systems as well as the design of efficient nanoscale organic optoelectronic devices.

A non-destructive n-doping method for graphene with precise control of electronic properties via atomic layer deposition.

Graphene applications require high precision control of the Fermi level and carrier concentration via a nondestructive doping method. Here, we develop an effective n-doping technique using atomic layer deposition (ALD) of ZnO thin films on graphene through a reactive molecular layer. This ALD doping method is nondestructive, simple, and precise. The ZnO thin films on graphene are uniform, conformal, of good quality with a low density of pinholes, and finely tunable in thickness with 1 Å resolution. We demonstrate graphene transistor control in terms of the Dirac point, carrier density, and doping state as a function of the ZnO thickness. Moreover, ZnO functions as an effective thin-film barrier against air-borne water and oxygen on the graphene, resulting in extraordinary stability in air for graphene devices. ZnO ALD was also applied to other two-dimensional materials including MoS2 and WSe2, which substantially enhanced electron mobility.

Inkjet-Assisted Nanotransfer Printing for Large-Scale Integrated Nanopatterns of Various Single-Crystal Organic Materials.

Inkjet-assisted nanotransfer printing (inkjet-NTP) facilitates spatial control of many arrays of various organic functional materials on a single substrate with a high-throughput integration process, enabling monolithic integration of various organic nanopatterns. Inkjet-NTP enables wafer-scale organic electronic circuits composed of field-effect transistors, complementary inverters, and p-n diodes, demonstrating its capability to produce a high-performance, multifunctional organic chip.

Correlation Between Corneal Button Size and Intraocular Pressure During Femtosecond Laser-Assisted Keratoplasty.

PURPOSE: To evaluate changes in intraocular pressure (IOP) in recipient and donor eyes during femtosecond laser-assisted keratoplasty (FLAK) and to assess for differences in the diameter of trephinated corneal buttons according to changes in pressure. MATERIALS AND METHODS: Twenty porcine whole eyes (recipient model) and 20 porcine-corneoscleral rims (donor model) were prepared, and anterior chamber pressures were measured using a fiberoptic sensing device (Opsens, Quebec, Canada) during the femtosecond laser corneal cutting process. To determine the diameter of corneal buttons, 10 porcine whole eyes (recipient model) and 12 corneoscleral rims (donor model) of each baseline IOP were cut with the femtosecond laser programmed to the following pattern: "vertical side cut"; 1200 µm (depth), 8 mm (diameter). Digital photographs were obtained using microscopy and subsequently analyzed. RESULTS: The IOP (mean ± SD) for the recipient model was 10.2 (±0.9) mm Hg at baseline and ranged from 96.6 (±4.5) to ∼138.4 (±3.8) mm Hg during the corneal cutting process. This shows that the maximum IOP during FLAK increased 13.5 times compared with baseline. In the donor model, the mean pressure elevation from baseline artificial anterior chamber (AAC) pressure to corneal cutting was 15.8 (±5.4) mm Hg. This showed a positive correlation with baseline IOP [correlation coefficient (CC) = 0.827, P = 0.006]. As the baseline IOP in the recipient eye increased, trephinated corneal button size was reduced by up to 3.9% in diameter (CC = -0.945, P = 0.015). In addition, in donor eyes, the diameter was decreased by up to 11.7% as the baseline AAC pressure increased (CC = -0.934, P = 0.006). CONCLUSIONS: During the FLAK procedure, the IOP increases in both recipient and donor eyes. The diameter of the trephinated donor and recipient corneal buttons was decreased as the initial baseline IOP increased. Ophthalmic surgeons can determine the AAC pressure based on the baseline IOP in the recipient patient.

Wafer-scale single-domain-like graphene by defect-selective atomic layer deposition of hexagonal ZnO.

Large-area graphene films produced by means of chemical vapor deposition (CVD) are polycrystalline and thus contain numerous grain boundaries that can greatly degrade their performance and produce inhomogeneous properties. A better grain boundary engineering in CVD graphene is essential to realize the full potential of graphene in large-scale applications. Here, we report a defect-selective atomic layer deposition (ALD) for stitching grain boundaries of CVD graphene with ZnO so as to increase the connectivity between grains. In the present ALD process, ZnO with a hexagonal wurtzite structure was selectively grown mainly on the defect-rich grain boundaries to produce ZnO-stitched CVD graphene with well-connected grains. For the CVD graphene film after ZnO stitching, the inter-grain mobility is notably improved with only a little change in the free carrier density. We also demonstrate how ZnO-stitched CVD graphene can be successfully integrated into wafer-scale arrays of top-gated field-effect transistors on 4-inch Si and polymer substrates, revealing remarkable device-to-device uniformity.

Assessment of Lens Center Using Optical Coherence Tomography, Magnetic Resonance Imaging, and Photographs of the Anterior Segment of the Eye.

PURPOSE: To determine the nearest marker for evaluating the center of the crystalline lens using optical coherence tomography (OCT), magnetic resonance imaging (MRI), and photographs. METHODS: Optical coherence tomography scans of human eyes were obtained in vivo during femtosecond laser-assisted cataract surgery. From axial and sagittal images, the distance of the angle center (AC) and pupil center (PC) from the scanned capsule center (SCC) was calculated. From pre- and postoperative photographs, the distance of the PC and limbal center (LC) from the intraocular lens (IOL) center was calculated, and distance between each center on the lens equatorial plane was compared. After combination of pre- and postoperative images, we arranged the centers in order of distance from the IOL center. High-resolution MRI was performed in pig eyes ex vivo to confirm the exact location of the lens center relative to other centers. RESULTS: In human OCT scans and photographs (n = 76), the IOL center to AC distance was 0.22 ± 0.13 mm, the IOL center to SCC distance was 0.22 ± 0.12 mm, the IOL center to PC distance was 0.25 ± 0.17 mm, and the IOL center to LC distance was 0.30 ± 0.18 mm. The AC and SCC were significantly closer to the IOL center than the PC or LC. In MRI (n = 54 images), the lens center to AC distance was 0.90 ± 0.58 mm, and the lens center to PC distance was 1.53 ± 0.87 mm (Δ distance = 0.63 ± 0.69 mm, P = 0.000). CONCLUSIONS: Optical coherence tomography, MRI, and photographs of the anterior segment revealed that the AC is the nearest marker to the center of the lens equator.

Open ring-shaped guider for circular continuous curvilinear capsulorhexis during cataract surgery.

UNLABELLED: We describe a new type of open-loop caliper for capsulorhexis during cataract surgery. This tool, which is made of poly(methyl methacrylate), can optimize capsulorhexis shape, size, and centration. One of the strengths of this tool, which derives from the open-loop design, is its ease of insertion and removal via a small incision site compared with other calipers used for capsulorhexis. The caliper is positioned on the anterior capsule after the anterior chamber is filled with an ophthalmic viscosurgical device and removed after creation of the continuous curvilinear capsulorhexis. The caliper enables the surgeon to achieve an ideal capsulorhexis and promotes long-lasting circularity. FINANCIAL DISCLOSURE: Dr. Joo is one of the inventors on the patent filed by the Catholic University of Korea covering details in this manuscript. Dr. Lee has no financial or proprietary interest in any material or method mentioned.

Cross-stacked single-crystal organic nanowire p-n nanojunction arrays by nanotransfer printing.

We fabricated cross-stacked organic p-n nanojunction arrays made of single-crystal 6,13-bis(triisopropylsilylethynyl) pentacene (TIPS-PEN) and fullerene (C60) nanowires as p-type and n-type semiconductors, respectively, by using a nanotransfer printing technique. Single-crystal C60 nanowires were synthesized inside nanoscale channels of a mold and directly transferred onto a desired position of a flexible substrate by a lubricant liquid layer. In the consecutive printing process, single-crystal TIPS-PEN nanowires were grown in the same way and then perpendicularly aligned and placed onto the C60 nanowire arrays, resulting in a cross-stacked single-crystal organic p-n nanojunction array. The cross-stacked single-crystal TIPS-PEN/C60 nanowire p-n nanojunction devices show rectifying behavior with on/off ratio of ∼ 13 as well as photodiode characteristic with photogain of ∼ 2 under a light intensity of 12.2 mW/cm(2). Our study provides a facile, solution-processed approach to fabricate a large-area array of organic crystal nanojunction devices in a desired arrangement for future nanoscale electronics.

Single-crystal poly(3,4-ethylenedioxythiophene) nanowires with ultrahigh conductivity.

We developed single-crystal poly(3,4-ethylenedioxythiopene) (PEDOT) nanowires with ultrahigh conductivity using liquid-bridge-mediated nanotransfer printing with vapor phase polymerization. The single-crystal PEDOT nanowires are formed from 3,4-ethylenedioxythiophene (EDOT) monomers that are self-assembled and crystallized during vapor phase polymerization process within nanoscale channels of a mold having FeCl3 catalysts. These PEDOT nanowires, aligned according to the pattern in the mold, are then directly transferred to specific positions on a substrate to generate a nanowire array by a direct printing process. The PEDOT nanowires have closely packed single-crystalline structures with orthorhombic lattice units. The conductivity of the single-crystal PEDOT nanowires is an average of 7619 S/cm with the highest up to 8797 S/cm which remarkably exceeds literature values of PEDOT nanostructures/thin films. Such distinct conductivity enhancement of single-crystal PEDOT nanowires can be attributed to improved carrier mobility in PEDOT nanowires. To demonstrate usefulness of single-crystal PEDOT nanowires, we fabricated an organic nanowire field-effect transistor array contacting the ultrahigh conductive PEDOT nanowires as metal electrodes.

Nano-photosensitizers Engineered to Generate a Tunable Mix of Reactive Oxygen Species, for Optimizing Photodynamic Therapy, Using a Microfluidic Device.

This work is aimed at engineering photosensitizer embedded nanoparticles (NPs) that produce optimal amount of reactive oxygen species (ROS) for photodynamic therapy (PDT). A revised synthetic approach, coupled with improved analytical tools, resulted in more efficient PDT. Specifically, methylene blue (MB) conjugated polyacrylamide nanoparticles (PAA NPs), with a polyethylene glycol dimethacrylate (PEGDMA, Mn 550) cross-linker, were synthesized so as to improve the efficacy of cancer PDT. The long cross-linker chain, PEGDMA, increases the distance between the conjugated MB molecules so as to avoid self-quenching of the excited states or species, and also enhances the oxygen permeability of the NP matrix, when compared to the previously used shorter cross-linker. The overall ROS production from the MB-PEGDMA PAA NPs was evaluated using the traditional way of monitoring the oxidation rate kinetics of anthracence-9,10-dipropionic acid (ADPA). We also applied singlet oxygen sensor green (SOSG) so as to selectively derive the singlet oxygen (1O2) production rate. This analysis enabled us to investigate the ROS composition mix based on varied MB loading. To effectively obtain the correlation between the ROS productivity and the cell killing efficacy, a microfluidic chip device was employed to provide homogeneous light illumination from an LED for rapid PDT efficacy tests, enabling simultaneous multiple measurements while using only small amounts of NPs sample. This provided multiplexed, comprehensive PDT efficacy assays, leading to the determination of a near optimal loading of MB in a PAA matrix for high PDT efficacy by measuring the light-dose-dependent cell killing effects of the various MB-PEGDMA PAA NPs using C6 glioma cancer cells.

Overcoming cancer multidrug resistance by codelivery of doxorubicin and verapamil with hydrogel nanoparticles.

The efficacy of chemotherapy is often inhibited by multidrug resistance (MDR). A highly engineerable hydrogel nanoparticle (NP) serves as a carrier for the optimal codelivery to tumor cells of the chemodrug, doxorubicin (Dox) and the chemosensitizer, verapamil (Vera), aiming at alleviating tumor MDR. The hydrogel NPs are prepared via the copolymerization of acrylamide and 2-carboxyethyl acrylate. Dox and Vera are post-loaded into the respective NPs, with drug loading around 7.7 wt% and 8.0 wt%, respectively. The codelivery of Dox-NPs and Vera-NPs increases the intracellular accumulation of Dox, and significantly enhances the cell killing ability of Dox with respect to NCI/ADR-RES cells in vitro. These findings suggest that such codelivery nanoplatforms provide a promising route for overcoming tumor MDR.