The retina offers a unique window into both ocular and systemic health, motivating the development of AI-based tools for disease screening and risk assessment. In this study, we present a comprehensive evaluation of six state-of-the-art deep neural networks, including convolutional neural networks and vision transformer architectures, on the Brazilian Multilabel Ophthalmological Dataset (BRSET), comprising 16,266 fundus images annotated for multiple clinical and demographic labels. We explored seven classification tasks: Diabetes, Diabetic Retinopathy (2-class), Diabetic Retinopathy (3-class), Hypertension, Hypertensive Retinopathy, Drusen, and Sex classification. Models were evaluated using precision, recall, F1-score, accuracy, and AUC. Among all models, the Swin-L generally delivered the best performance across scenarios for Diabetes (AUC = 0.88, weighted F1-score = 0.86), Diabetic Retinopathy (2-class) (AUC = 0.98, weighted F1-score = 0.95), Diabetic Retinopathy (3-class) (macro AUC = 0.98, weighted F1-score = 0.95), Hypertension (AUC = 0.85, weighted F1-score = 0.79), Hypertensive Retinopathy (AUC = 0.81, weighted F1-score = 0.97), Drusen detection (AUC = 0.93, weighted F1-score = 0.90), and Sex classification (AUC = 0.87, weighted F1-score = 0.80). These results reflect excellent to outstanding diagnostic performance. We also employed gradient-based saliency maps to enhance explainability and visualize decision-relevant retinal features. Our findings underscore the potential of deep learning, particularly vision transformer models, to deliver accurate, interpretable, and clinically meaningful screening tools for retinal and systemic disease detection.

In this paper, we develop an analytical approach by linearization of small-signal rate equations in the presence of the Langevin noise sources to investigate the noise properties of mid-IR optically injected locked quantum cascade lasers. Excellent agreement between the reported numerical results and our developed analytical method have been obtained with the advantage of drastic calculation time reduction and deeper system performance knowledge. Then, a comprehensive study has been performed which reveals notable noise reduction, especially near the threshold current. It is shown that increasing the phase difference between the slave and master lasers enhances the corresponding noise term and makes this term the dominant noise-generating process near the locking edges. Additionally, by introducing relative intensity noise (RIN) maps, we demonstrate that increasing the bias current leads to lower RIN and frequency detuning range, while increasing the optical injection power eventuates wider frequency range and also, lower RIN. It has also been discussed that higher linewidth enhancement factors enhance the locking range and decreases the RIN near the positive frequency detuning boundaries. RIN reduction efficiency versus total injected power to the slave laser has been also introduced as a measure of effectiveness of optical injection locking.

Relative intensity noise (RIN) behavior of a quantum dot laser (QDL) at free running and under optical injection locking (OIL) has been reported in detail. Considering the mutual effects of the inhomogeneous and homogeneous broadenings, different and also non-routine RIN characteristics of a QDL at low, intermediate, and high homogeneous broadening regimes have been discussed. Results demonstrate that an injection-locked quantum dot laser exhibit enhanced characteristic over free-running operation along with low-frequency noise behavior in a wide range of homogeneous broadening values. RIN reduction of about 25, 55, and 7 dB/Hz have been achieved for low, intermediate, and high values of homogeneous broadening which implies the spontaneous emission noise suppression even at low injection power. The promising impact of the OIL on side-mode suppression for the intermediate homogeneous broadening regime has also been discussed. Our results guarantee the single-mode and low-noise operation of the OIL-QDL at any homogeneous broadening value.

Quantum cascade lasers with quantum dot (QD) active regions have been developed to overcome the drawbacks of quantum-well-based cascade lasers. Low threshold current and enhanced bandwidth characteristics of QD lasers can be combined with promising energy-efficient features of vertical cavity surface emitting lasers (VCSELs) through introducing the novel, to the best of our knowledge, concept of a QD cascade VCSEL (QDC-VCSEL). Design steps and predicted features of a QDC-VCSEL operating at 4.5 µm wavelength are discussed in this paper, where the cavity and active region are optimized for TE mode using a high contrast grating design. Laser characteristics are investigated by developing rate equations for the cascade scheme. Threshold and L-I characteristics are obtained at different temperatures. Also, the dependency of threshold current, slope efficiency, and modulation bandwidth on design parameters, mainly the carrier tunneling time, is studied. Results exhibit output peak powers of ∼ 2 and ∼ 0.5 m W at 150 and 273 K, respectively. In addition, threshold current increment and modulation bandwidth reduction are observed with the increase in tunneling time.

A topological laser array based on the non-Hermitian Su-Schrieffer-Heeger chain of coupled microring resonators has been presented in this manuscript. Based on the theory of the active and passive microrings in each unit cell, we have defined a PT-symmetry structure where the edge states can be protected and the imaginary gauge field cause delocalize light between the microrings. Numerical simulations based on the rate equation model and stability analysis have been performed to demonstrate the stability criteria of the presented structure. Utilizing this configuration, it has been shown that the number of protected topological supermodes can be controlled by the number of active and passive microrings in each unit cell while all the other extended supermodes are transformed into the non-topological states. Design flexibility to expand and modify the lasing stability region and topological supermode selection can be obtained based on the introduced configuration.

Temperature-dependent characteristics of a midinfrared quantum cascade laser (QCL) including the threshold current, the output power, the modulation bandwidth, and the relative intensity noise (RIN) exhibit a deterioration with temperature from 77 to 300 K. Based on a comprehensive temperature-dependent model we show that optical injection locking can be applied to surpass the intrinsically limited thermal characteristics of a QCL through the modification of the carrier density in the active region. A modulation bandwidth improvement of $\approx 35\%$ and an almost temperature-independent RIN behavior for an injection-locked laser have been achieved compared to a free-running slave laser at room temperature. A threshold current reduction of $\approx 20\%$ has also been obtained at 300 K. Results indicate that the locking boundaries of an injection-locked QCL are suppressed with temperature. The output power and the small-signal characteristics of the laser have also been analyzed in the entire stable locking range.