Template-based Identification of Non-malignant Skin lesions to Minimize Biopsy: Experiments on Hundred Images from ISIC Archive

Rai Sachindra Prasad, Vikas Prasad

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

2 Scopus citations

Abstract

Diagnosis of skin cancer is based on clinical examination and biopsy. The objective of this study is to propose a simple and quick non-invasive image processing-based diagnostic method which identifies non-malignant skin lesions in order to minimize biopsy to the maximum possible extent. This novel study proposes selection of a master template (MT) of a pre-diagnosed malignant dermatoscopic image with histopathology report from ISIC archive. Identification of non-malignant image is implemented using scatter plots of chromatic parameter Saturation (S) in HSV space on the subtracted image from the MT. Two thresholds of S, 0.45 for non-malignant, and 0.35 for malignant lesions, in the scatter plots are the identifiers. The hypothesis of the proposed method is that the color and intensity hold the key for identification of non-malignant lesions. Experiments on hundred images from ISIC archive achieved an accuracy of 80%, while the specificity of 85% and sensitivity of 71% were obtained.

Original languageEnglish
Title of host publication2021 9th E-Health and Bioengineering Conference, EHB 2021
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781665440004
DOIs
StatePublished - 2021
Event9th IEEE International Conference on E-Health and Bioengineering Conference, EHB 2021 - Iasi, Romania
Duration: Nov 18 2021Nov 19 2021

Publication series

Name2021 9th E-Health and Bioengineering Conference, EHB 2021

Conference

Conference9th IEEE International Conference on E-Health and Bioengineering Conference, EHB 2021
Country/TerritoryRomania
CityIasi
Period11/18/2111/19/21

Keywords

  • HSV
  • Malignant skin lesions
  • Spectral analyses

Fingerprint

Dive into the research topics of 'Template-based Identification of Non-malignant Skin lesions to Minimize Biopsy: Experiments on Hundred Images from ISIC Archive'. Together they form a unique fingerprint.

Cite this