Scholarlink

<p>Teledermoscopy (TDS) offers an efficient alternative to face-to-face assessment of suspicious skin lesions. Although diagnostic accuracy in TDS is relatively high, further improvement is desirable to reduce the risk of overlooked skin cancers and unnecessary interventions of benign lesions. Image enhancement (IE), i.e. digital modification of different parameters of an image, may potentially improve visualisation and thereby diagnostic accuracy. Several tools for IE were reviewed through two workshops and two were eventually selected for the study: <i>Contrast Limited Adaptive Histogram Equalisation</i> (CLAHE) for contrast enhancement and <i>Gamma Correction</i> for brightness adjustment. A retrospective randomized clinical study was performed in which 1997 TDS cases were distributed to 13 assessors with varying dermoscopic experience, who assessed lesions either with access to IE-tools (IE-group) or without (control group). Diagnostic accuracy for malignant vs. benign and melanoma-group vs. non-melanoma-group lesions was compared using chi-square tests. A generalized linear mixed model (GLMM) with assessor as a random intercept and adjustment for image quality was applied to account for assessor-dependent variance. Management recommendations were analyzed using both graded intervention levels relative to a gold standard and a stricter harm–benefit framework, assessing inappropriately under-management of malignant and over-management of benign lesions. No significant differences were found between the groups regarding malignant/benign classification. For melanoma differentiation, specificity was slightly, but significantly lower in the IE group compared to the control group (86.3; 95% confidence interval (CI) 83.8–88.5 vs. 89.9%; 95% CI 87.6–91.8, respectively, <i>p</i> = 0.02) while sensitivity remained unchanged. GLMM analyses confirmed these findings. No significant differences were observed in appropriateness of management recommendations or diagnostic confidence. Use of the IE tools <i>CLAHE</i> and <i>Gamma Correction</i> did not improve diagnostic performance in TDS assessments. These findings do not support routine clinical use of these IE tools for the purpose of improving diagnostic assessment of skin lesions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Image enhancements by CLAHE and gamma correction do not improve diagnostic accuracy in teledermoscopy: a retrospective cohort study

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月14日

查看全文
  • Hedvig Kristensson,
  • Carolina Nätterdahl,
  • Bertil Persson,
  • Karim Saleh,
  • Johan Kappelin,
  • Teo Helkkula,
  • Fredrik Johansson,
  • Johan Palmgren,
  • Jan Lapins,
  • Niki Rados,
  • Lina U. Ivert,
  • Karina Schultz,
  • Kari Nielsen,
  • Åsa Ingvar

摘要

Teledermoscopy (TDS) offers an efficient alternative to face-to-face assessment of suspicious skin lesions. Although diagnostic accuracy in TDS is relatively high, further improvement is desirable to reduce the risk of overlooked skin cancers and unnecessary interventions of benign lesions. Image enhancement (IE), i.e. digital modification of different parameters of an image, may potentially improve visualisation and thereby diagnostic accuracy. Several tools for IE were reviewed through two workshops and two were eventually selected for the study: Contrast Limited Adaptive Histogram Equalisation (CLAHE) for contrast enhancement and Gamma Correction for brightness adjustment. A retrospective randomized clinical study was performed in which 1997 TDS cases were distributed to 13 assessors with varying dermoscopic experience, who assessed lesions either with access to IE-tools (IE-group) or without (control group). Diagnostic accuracy for malignant vs. benign and melanoma-group vs. non-melanoma-group lesions was compared using chi-square tests. A generalized linear mixed model (GLMM) with assessor as a random intercept and adjustment for image quality was applied to account for assessor-dependent variance. Management recommendations were analyzed using both graded intervention levels relative to a gold standard and a stricter harm–benefit framework, assessing inappropriately under-management of malignant and over-management of benign lesions. No significant differences were found between the groups regarding malignant/benign classification. For melanoma differentiation, specificity was slightly, but significantly lower in the IE group compared to the control group (86.3; 95% confidence interval (CI) 83.8–88.5 vs. 89.9%; 95% CI 87.6–91.8, respectively, p = 0.02) while sensitivity remained unchanged. GLMM analyses confirmed these findings. No significant differences were observed in appropriateness of management recommendations or diagnostic confidence. Use of the IE tools CLAHE and Gamma Correction did not improve diagnostic performance in TDS assessments. These findings do not support routine clinical use of these IE tools for the purpose of improving diagnostic assessment of skin lesions.