MUMBAI, India, June 16 -- Intellectual Property India has published a patent application (202611053430 A) filed by Ms. Prachi Pundhir; Mr. Gaurav Gupta; Mr. Saurabh; and Mr. Mohit, Ghaziabad, Uttar Pradesh, on April 27, for 'skin disease detection using dermoscopic image.'
Inventor(s) include Ms. Prachi Pundhir; Mr. Gaurav Gupta; Mr. Saurabh; and Mr. Mohit.
The application for the patent was published on June 5, under issue no. 23/2026.
According to the abstract released by the Intellectual Property India: "Skin cancer, particularly melanoma, kills thousands every year - yet it is one of the most treatable cancers when found early. The problem is not treatment. It is late detection, mostly because skilled dermatologists are scarce, especially in rural areas. This work builds an automated classification system that takes a dermoscopic image as input and tells you whether the lesion is benign or malignant. Before anything goes into the model, the image is cleaned - hair is removed using black hat morphological filtering, contrast is fixed with CLAHE, and the lesion is isolated using Otsu thresholding. After that, a pretrained ResNet50 network pulls a 2048-dimensional feature vector from each image. That vector goes into a Random Forest classifier with 200 trees for the final call. The system was tested on 8,000 images from the ISIC dataset - 6,400 for training, 1,600 held out for testing. Random Forest hit 86.7% accuracy, 84.8% precision, 85.3% recall, and an F1-score of 85.0%. The whole thing runs as a Streamlit."
Disclaimer: Curated by HT Syndication.
