Chennai Mathematical Institute

Skin diseases affect a considerable part of the world population, but it is not easy to diagnose these diseases on time because of the limited availability of dermatologists and the similarity of lesions. This research investigates advanced image processing techniques by combining OpenCV—Python library and EfficientNet—B3 architecture in order to conduct accurate and reliable classification of skin diseases. The model combines lesion-focused pre-processing and imbalance data augmentation for better feature learning. The results of the proposed approach reached 92.6% accuracy, which was much higher than that of traditional architectures such as ResNet-50 (83%), VGG-16 (82%), and InceptionV3 (87%). These results mention the effectiveness of the compound scaling for dermatological image analysis. Clinically, the system can be used for early detection and dermatology screening as it provides a fast automated screening, especially in parts of the world with a shortage of specialists. This plays a role in better triaging and lessening the time of diagnostic delay, as well as providing more accessible dermatological care in remote populations. Overall, the proposed method provides a reliable and efficient solution for the scalable classification of skin diseases.

Biomolecular condensates formed via liquid-liquid phase separation (LLPS) are essential for cellular organization. α-Synuclein, an amyloidogenic protein linked to Parkinson’s Disease (PD), undergoes phase separation at high concentrations, but the influence of lipid membranes on this process remains unclear. Here, combining in vitro reconstitution, cell biology, and simulations, we show that membranous interfaces promote α-Synuclein condensation at physiologically relevant sub-critical concentrations (~10 nM) without crowding agents. Notably, condensation occurs only on membranes with a specific stoichiometry of lipids, underscoring the role of interfacial potential. These condensates serve as nucleation sites for fibril formation, leading to membrane deformation and rupture. A lattice gas model reveals this behavior as a prewetting-like transition, where an attractive membrane induces local phase separation below the bulk saturation concentration. Indeed altering interfacial potential by lipid composition and membrane depolarization not only drastically changes α-Synuclein puncta size and number but also triggers their release from neurons. These findings reveal the crucial role of lipid membrane interfaces in regulating α-Synuclein condensation, aggregation and release, shedding light on a potential mechanism of their cell-to-cell propagation during neurodegeneration.

Brain tumor segmentation from MRI images is a critical task for accurate diagnosis and treatment planning. While traditional segmentation approaches leverage deep learning models, such as U-Net and its variants, they often struggle with balancing precision and interpretability, especially in handling noisy and complex MRI data. Earlier works have demonstrated remarkable performance improvements through hybrid models, yet limitations persist in effectively capturing both global and local contexts, as well as ensuring model transparency for clinical application. These shortcomings highlight the need for frameworks that not only excel in segmentation accuracy but also provide explainability for reliable medical decision-making. To address these challenges, we propose a novel Explainable U-Net Transformer (ExU-Trans) framework for brain tumor segmentation in MRI images. ExU-Trans integrates the interpretability strengths of Vision Transformers with the spatial precision of U-Net in a dual-encoder structure. The Vision Transformer encoder employs Self-Explanatory Multi-Head Attention (SE-MHA) to generate noise-resistant attention maps, while a modified U-Net encoder focuses on extracting multiscale spatial features critical for tumor delineation. To enhance interpretability, a Discriminative Attribute Explainer (DAE) is embedded within the transformer encoder, pinpointing tumor-specific attributes at a pixel level. Furthermore, a Contextual Self-Attention mechanism enriches feature representations by capturing both global and local contexts, refining tumor detection. The outputs from both encoders are combined using a bivariate fusion module that incorporates spatial and channel attention, producing a unified feature representation. Finally, an attribute-guided loss function enables self-supervised learning, resulting in highly accurate and inherently interpretable segmentation outcomes. Extensive evaluations demonstrate the superiority of ExU-Trans in terms of segmentation accuracy and explainability, offering a reliable tool for clinical brain tumor diagnosis and treatment planning.

Mobile Ad Hoc Networks (MANETs) are highly dynamic and resource-constrained systems in which the resourceful maintenance of secure, energy-efficient, and trustworthy communication is challenging owing to frequent changes in topologies and highly malicious intrusions. To resolve these issues, this work proposes research will provide a Hybrid Intelligence-Powered Secure Clustering and Trust-Optimized Routing (HISCTR) framework that combines adaptive clustering, intelligent trust assessment, and hybrid deep-learning-based intrusion detection into a single architecture. The Geographical Adaptive Fidelity Clustering (GAFC) scheme provides stable clustering, whereas the Addax Optimization Algorithm (AOA) identifies energy-balanced Cluster Heads and Sub-Cluster Heads according to residual energy, centrality, and mobility. Fuzzy Trust Vector Computation (FTVC) and Halfway Escape Optimization (HEO) are further used to provide routing reliability, as they collaboratively determine low-delay and highly trusted communication paths. Transformer-based temporal learning and CNN-based spatial analysis allow for a dual-path anomaly detection module to ensure the complete detection of multiple attacks. The proposed model showed the best performance at a mean power use of 0.8mJ, throughput of 0.92 Mbps, packet delivery ratio, 5300-round network life, and efficiency rate of 99.5%, as proven by extensive simulations. These findings verify that HISCTR can provide a security-conscious, scalable, and robust communication system for next-generation MANETs.

A catalytic Turing machine is a model of computation that is created by equipping a Turing machine with an additional auxiliary tape, which is initially filled with arbitrary content; the machine can read or write on the auxiliary tape during the computation, but it is constrained to halt with the same content in the auxiliary tape as it initially had. This article, studies the power of some natural variants of catalytic Turing machines with \(O(\log n)\) -size work tape and a polynomial-size auxiliary tape. We first define the notion of unambiguous catalytic Turing machines and prove that under a standard derandomization assumption, the class of problems solved by unambiguous catalytic Turing machines is the same as the class of problems solved by nondeterministic catalytic Turing machines. We then introduce the notion of randomized catalytic Turing machines and show that the resulting complexity class \(\mathsf {CBPL}\) is contained in the class \(\mathsf {ZPP}\) . We also explore the notion of symmetricity in the context of catalytic computation and prove that, under the same assumption as before, randomized catalytic Turing machines, symmetric catalytic Turing machines, and deterministic catalytic Turing machines that run in polynomial time are equally powerful.