DDCNN‑F: double decker convolutional neural network ’F’ feature fusion as a medical image classification framework

Sep 21, 2026

Upload by Jiven

Melanoma, a severe skin cancer, requires early detection to prevent rapid metastasis. This study presents a new feature-fusion framework for melanoma classification using a novel ‘F’ Flag feature for early identification of malignant versus benign skin lesions. A Double Decker Convolutional Neural Network (DDCNN) integrates a hair-removed, augmented Baseline Separated Channel with bottleneck features and ABCDE clinical bio-indicators, then evaluates hybrid fused features using multiple classifiers. Trained on ISIC 2019 and ISIC 2020, results report specificity 98.4%, accuracy 93.75%, precision 98.56%, and AUC 0.98, with higher classifier performance enabled by the ‘F’ indicator.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages24

Reading Time1 h 0 min

AccessFree to download

DDCNN‑F Feature Fusion Framework

Problem Background and Existing Approaches

Proposed Double Decker CNN Architecture

‘F’ Flag Feature and Hybrid Feature Fusion

Datasets and Experimental Results

[www. nature.com/scientificreports](www. nature.com/scientificreports) OPEN DDCNN‑F: double decker convolutional neural network ’F’feature fusion as a medical image classification framework Nirmala Veeramani1, Premaladha Jayaraman1*, Raghunathan Krishankumar2, Kattur Soundarapandian Ravichandran3 & Amir H. Gandomi4,5* Melanoma is a severe skin cancer that involves abnormal cell development. This study aims to provide a new feature fusion framework for melanoma classification that includes a novel ‘F’Flag feature for early detection. This novel ‘F’ indicator efficiently distinguishes benign skin lesions from malignant ones known as melanoma. The article proposes an architecture that is built in a Double Decker Convolutional Neural Network called DDCNN future fusion. The network’s deck one, known as a Convolutional Neural Network (CNN), finds difficult‑to‑classify hairy images using a confidence factor termed the intra‑class variance score. These hirsute image samples are combined to form a Baseline Separated Channel (BSC). By eliminating hair and using data augmentation techniques, the BSC is ready for analysis. The network’s second deck trains the pre‑processed BSC and generates bottleneck features. The bottleneck features are merged with features generated from the ABCDE clinical bio indicators to promote classification accuracy. Different types of classifiers are fed to the resulting hybrid fused features with the novel ’F’ Flag feature. The proposed system was trained using the ISIC 2019 and ISIC 2020 datasets to assess its performance. The empirical findings expose that the DDCNN feature fusion strategy for exposing malignant melanoma achieved a specificity of 98.4%, accuracy of 93.75%, precision of 98.56%, and Area Under Curve (AUC) value of 0.98. This study proposes a novel approach that can accurately identify and diagnose fatal skin cancer and outperform other state‑of‑ the‑art techniques, which is attributed to the DDCNN‘F’ Feature fusion framework. Also, this research ascertained improvements in several classifiers when utilising the ‘F’ indicator, resulting in the highest specificity of + 7.34% . Skin cancer is caused by abnormal cells and can metastasise quickly if not detected early1. Of the number of studies on melanoma using different analytical platforms, one of the pioneering efforts is Computer-Aided Diagnosis (CAD)2. These research efforts used automated algorithms to identify cutaneous lesions upfront by analysing image colour, shape, and structure3,4. Image segmentation is still a prominent field of study—the progression of segmentation by statistical analysis of medical images5. Extraction of characteristics from the lesion region assists in categorisation, although developing such systems has been challenging and sluggish6. The majority of automated detection systems use the Asymmetry, Border, Color, Diameter, and Evolution (ABCDE) rule. In 2001, Senan et al. were the first to work with the four aspects of this technique, which involved 122 parameters, and used the conventional models with an automation system employed7. Squamous-cell carcinoma (SCC), basal-cell carcinoma (BCC), and melanoma are the three kinds of skin cancer8; the former two are less prevalent types and are grouped as non-melanoma skin cancer. While BBC develops slowly, possibly causing tissue damage, it seldom causes mortality and is usually seen in the nasal cavity9, 10. Squamous-cell skin cancer manifestsas a hard mass with a scaly surface and crust, which may lead to the formation of an ulcer11. Melanoma is the deadliest and most severe type of skin cancer that can metastasise and become chronic and life-threatening if 1School of Computing, SASTRA Deemed to Be University, Thanjavur, India. 2Information Technology Systems and Analytics Area, Indian Institute of Management Bodh Gaya, Bodh Gaya, Bihar 824234, India. 3Department of Mathematics, Amrita School of Physical Sciences, Amrita Vishwa Vidyapeetham, Coimbatore, India

FAQs

What is the goal of the DDCNN‑F framework for melanoma classification?

The framework aims to improve melanoma detection by fusing features using a novel ‘F’ Flag feature to distinguish benign lesions from malignant melanoma for earlier diagnosis.

How does the proposed double decker network work?

The first deck uses a CNN to handle difficult-to-classify hairy images and forms a Baseline Separated Channel after hair removal and data augmentation. The second deck trains on the pre-processed channel to produce bottleneck features that are merged with ABCDE clinical bio-indicator features.

Which performance metrics were achieved and on what datasets?

The system was trained and evaluated on ISIC 2019 and ISIC 2020 datasets, reaching specificity 98.4%, accuracy 93.75%, precision 98.56%, and AUC 0.98.

Sep 21, 2026

Upload by Jiven