MUMBAI, India, April 17 -- Intellectual Property India has published a patent application (202641043792 A) filed by Dr. Ganeshmoorthy S; Deepika S; Meenaloshini V; Sriram D; Sutharssan S; Vishwa A; Vijaya Sri S; Srimathi P; and Abhishek P, Coimbatore, Tamil Nadu, on April 6, for 'quantum-classical hybrid learning for multi-class skin disease detection with resource optimization.'
Inventor(s) include Dr. Ganeshmoorthy S; Deepika S; Meenaloshini V; Sriram D; Sutharssan S; Vishwa A; Vijaya Sri S; Srimathi P; and Abhishek P.
The application for the patent was published on April 17, under issue no. 16/2026.
According to the abstract released by the Intellectual Property India: "Quantum-Classical Hybrid Learning for Multi-Class Skin Disease Detection with Resource Optimization" Abstract: Skin disease classification remains a challenging problem in modern dermatology due to the high visual similarity among different disease categories, often leading to diagnostic ambiguity even for experienced clinicians (Esteva et al., 2017). Although classical deep learning models, particularly convolutional neural networks (CNNs), have demonstrated strong performance in medical image analysis (LeCun et al., 2015), they are limited by their dependence on large, balanced datasets and high computational requirements (Litjens et al., 2017). To overcome these limitations, this study proposes a resource-efficient hybrid quantum- classical framework that integrates deep feature extraction using ResNet50 with enhanced pattern recognition capabilities of quantum machine learning (Cerezo et al., 2021). The proposed model is implemented using PyTorch 2.0 and Qiskit 0.43 and achieves a validation accuracy of 87.33% and a test accuracy of 87.2% while utilizing only 100 training samples per class. This represents a significant reduction of approximately 90% in training data requirements compared to conventional deep learning approaches (Aravinda et al., 2023). Furthermore, the hybrid model demonstrates a substantial reduction in computational complexity, utilizing only 32 trainable quantum parameters compared to approximately 25.5 million parameters in traditional CNN architectures, resulting in a 99.87% reduction in parameter count (Schuld & Petruccione, 2018). The model also exhibits high computational efficiency, completing training in approximately 245 seconds on a standard CPU, which is significantly faster than classical training pipelines. The system is deployed using FastAPI with an average inference latency of 519 ms, demonstrating its practical applicability in real- time clinical environments. Skin diseases are among the most prevalent health conditions worldwide, affecting millions of individuals across different age groups and geographical regions. These conditions range from mild infections to severe chronic disorders, including melanoma, psoriasis, and eczema, which may significantly impact an individual's quality of life if not diagnosed at an early stage (Esteva et al., 2017). Early and accurate diagnosis of skin diseases is therefore essential for effective treatment planning and prevention of complications. However, the visual similarity among various dermatological conditions, characterized by overlapping features such as color variations, texture patterns, and lesion shapes, makes manual diagnosis a complex and error-prone task even for experienced dermatologists (Codella et al., 2019). With the rapid advancement of artificial intelligence, automated skin disease classification has emerged as a promising solution to support clinical decision-making. In particular, deep learning techniques, especially Convolutional Neural Networks (CNNs), have demonstrated remarkable success in image classification tasks, including medical imaging applications (LeCun et al., 2015). Architectures such as ResNet, VGGNet, and DenseNet have been widely adopted due to their ability to automatically learn hierarchical feature representations from raw image data, eliminating the need for manual feature engineering (He et al., 2016). These models have achieved dermatologist-level performance in certain skin cancer classification tasks when trained on large annotated datasets (Esteva et al., 2017). combination allows the system to leverage the strengths of both paradigms-efficient feature learning from classical models and enhanced pattern representation from quantum circuits (Cerezo et al., 2021). Recent studies suggest that hybrid models can achieve competitive performance using significantly fewer parameters and reduced training data, making them suitable for real-world applications in resource- constrained environments (Schuld & Petruccione, 2018). Despite these advancements, classical deep learning approaches face several significant challenges. One of the primary limitations is their strong dependency on large-scale labeled datasets, which are often difficult to obtain in the medical domain due to privacy concerns, limited availability of expert annotations, and high labeling costs (Litjens et al., 2017). Additionally, deep neural networks typically involve millions of trainable parameters, resulting in high computational complexity, increased training time, and the need for specialized hardware such as GPUs or TPUs. These constraints limit the practical deployment of such models in resource- constrained environments, including small clinics and edge devices. To address these limitations, recent research has explored the integration of quantum computing with machine learning, leading to the emergence of Quantum Machine Learning (QML). Quantum computing leverages fundamental principles such as superposition, entanglement, and quantum interference to perform computations in high-dimensional Hilbert spaces, potentially offering advantages over classical approaches for certain complex problems (Biamonte et al., 2017). Although large-scale fault-tolerant quantum computers are still under development, the current Noisy Intermediate-Scale Quantum (NISQ) era provides an opportunity to explore hybrid quantum-classical algorithms that combine classical neural networks with parameterized quantum circuits (Preskill, 2018)."
Disclaimer: Curated by HT Syndication.
