Numerical simulation of ultrawideband supercontinuum generation covering a 2-20 µm waveband in cascaded all-soft-glass fiber.

We designed a cascaded all-soft-glass fiber structure and simulate midinfrared 2-20 µm ultrawideband supercontinuum (SC) generation numerically. The cascaded fiber structure consists of a 1.5 m I n F 3 fiber, a 0.2 m chalcogenide photonic crystal fiber, and a 0.2 m tellurium-based chalcogenide photonic crystal fiber. Using a 2 µm pulse pumping this cascaded structure, the generated SC covering the wavelengths longer than 20 µm has been demonstrated theoretically. The 30 dB bandwidth reaches 20.87 µm from 1.44 to 22.31 µm. The effect of different pulse widths on SC generation is considered. With the increase of peak power and the decrease of pulse width, the energy of SC in the 15-20 µm waveband increases gradually. The mechanism of SC broadening process has also been analyzed. The SC generation of more than 20 µm in this cascade structure is caused by the self-phase modulation, soliton effects, four-wave mixing, and redshifted dispersive wave. This method demonstrates the possibility of generating ultrawide bandwidth SCs up to a 20 µm waveband by a commercial 2 µm pump source and all-fiber structure.

Decaying dynamics of harmonic mode-locking in a SESAM-based mode-locked fiber laser.

The entire decaying dynamics of harmonic mode-locking (HML) are studied utilizing the dispersive Fourier transform (DFT) technique in a SESAM-based mode-locked fiber laser. It is unveiled that the harmonic solitons do not disappear directly, but undergo transitional processes from the higher-order HML to the lower-order HML and then to the fundamental mode-locking (FML), and finally vanish. The "big corner" can also exist in the decaying process rather than just in the buildup process of HML, and there is at least one "big corner" during the decaying process between the consecutive multi-pulsing states. The energy stabilization phase (ESP) cannot be observed during every transitional process. A breathing behavior and a vibrating soliton molecule are observed in the decaying process from the 2nd HML to the FML and in the decaying process of the FML, respectively. Our work would enrich the understanding of HML behaviors and may contribute to the laser designs.

High-efficiency mode-locked erbium-doped ZBLAN fiber laser around 2.8 µm by directly depositing Bi2S3 particles onto a cavity mirror.

Mid-infrared (MIR) pulsed lasers near a 3 µm waveband show great potential for the high absorption of water molecules and many important gas molecules. A passively Q-switched mode-locked (QSML) E r 3+-doped fluoride fiber laser with a low laser threshold and high slope efficiency around a 2.8 µm waveband is reported. The improvement is achieved by depositing bismuth sulfide (B i 2 S 3) particles onto the cavity mirror directly as a saturable absorber and using the cleaved end of the fluoride fiber as output directly. -QSML pulses begin to appear with the pump power of 280 mW. The repetition rate of the QSML pulses reaches a maximum of 33.59 kHz with the pump power of 540 mW. When the pump power is further increased, the output of the fiber laser switches from the QSML to the continuous-wave mode-locked operation with the repetition rate of 28.64 MHz and the slope efficiency of 12.2%. The results indicate that B i 2 S 3 is a promising modulator for the pulsed lasers near a 3 µm waveband, which paves the way for further development of various applications in MIR wavebands, including material processing, MIR frequency combs, and modern healthcare.

Medical image segmentation with generative adversarial semi-supervised network.

Recent medical image segmentation methods heavily rely on large-scale training data and high-quality annotations. However, these resources are hard to obtain due to the limitation of medical images and professional annotators. How to utilize limited annotations and maintain the performance is an essential yet challenging problem. In this paper, we try to tackle this problem in a self-learning manner by proposing a generative adversarial semi-supervised network. We use limited annotated images as main supervision signals, and the unlabeled images are manipulated as extra auxiliary information to improve the performance. More specifically, we modulate a segmentation network as a generator to produce pseudo labels for unlabeled images. To make the generator robust, we train an uncertainty discriminator with generative adversarial learning to determine the reliability of the pseudo labels. To further ensure dependability, we apply feature mapping loss to obtain statistic distribution consistency between the generated labels and the real labels. Then the verified pseudo labels are used to optimize the generator in a self-learning manner. We validate the effectiveness of the proposed method on right ventricle dataset, Sunnybrook dataset, STACOM, ISIC dataset, and Kaggle lung dataset. We obtain 0.8402-0.9121, 0.8103-0.9094, 0.9435-0.9724, 0.8635-0.886, and 0.9697-0.9885 dice coefficient with 1/8 to 1/2 proportion of densely annotated labels, respectively. The improvements are up to 28.6 points higher than the corresponding fully supervised baseline.

Cardiac MRI segmentation with focal loss constrained deep residual networks.

Delineating anatomical structures for cardiac magnetic resonance imaging (CMRI) is crucial for various medical applications such as medical diagnoses, treatment, and pathological studies. CMRI segmentation, which aims to automatically and accurately segment the heart structures, is highly beneficial for cardiologists. However, it is non-trivial to perfectly segment the ventricles, especially for the heart apex slices, considering their small sizes compared to the input images. For example, the endocardium in the Sunnybrook dataset only occupies 4% of the entire image by average. During the training process, these target pixels, or other hard samples, are buried by the massive backgrounds that make the model mostly receive optimization signals from easy samples. In this paper, we propose a focal loss constrained residual network (FR-Net) to tackle the problem. In order to mitigate the fact that the gradients of the hard samples can be easily overwhelmed by the easy samples, we use a pixel-wise re-weighting strategy to balance the gradients. Furthermore, considering focal loss constraints for each pixel independently, we propose an alternative training fashion that trains the model with focal loss and dice loss alternatively. The segmentation model can not only benefit from the pixel-wise focal loss but also from the region-wise dice loss to comprehensively optimize the model. We conducted thorough experiments on the Sunnybrook dataset, CMRI dataset, right ventricle dataset, and ACDC dataset to verify the effectiveness of the proposed method.