TY - JOUR
T1 - Automated grading system for quantifying KOH microscopic images in dermatophytosis
AU - KV, Rajitha
AU - Govindan, Sreejith
AU - PY, Prakash
AU - Kamath, Asha
AU - Rao, Raghavendra
AU - Prasad, Keerthana
N1 - Publisher Copyright:
© 2024 The Authors
PY - 2025/1
Y1 - 2025/1
N2 - Concerning the progression of dermatophytosis and its prognosis, quantification studies play a significant role. Present work aims to develop an automated grading system for quantifying fungal loads in KOH microscopic images of skin scrapings collected from dermatophytosis patients. Fungal filaments in the images were segmented using a U-Net model to obtain the pixel counts. In the absence of any threshold value for pixel counts to grade these images as low, moderate, or high, experts were assigned the task of manual grading. Grades and corresponding pixel counts were subjected to statistical procedures involving cumulative receiver operating characteristic curve analysis for developing an automated grading system. The model's specificity, accuracy, precision, and sensitivity metrics crossed 92%, 86%, 82%, and 76%, respectively. ’Almost perfect agreement’ with Fleiss kappa of 0.847 was obtained between automated and manual gradings. This pixel count-based grading of KOH images offers a novel, cost-effective solution for quantifying fungal load.
AB - Concerning the progression of dermatophytosis and its prognosis, quantification studies play a significant role. Present work aims to develop an automated grading system for quantifying fungal loads in KOH microscopic images of skin scrapings collected from dermatophytosis patients. Fungal filaments in the images were segmented using a U-Net model to obtain the pixel counts. In the absence of any threshold value for pixel counts to grade these images as low, moderate, or high, experts were assigned the task of manual grading. Grades and corresponding pixel counts were subjected to statistical procedures involving cumulative receiver operating characteristic curve analysis for developing an automated grading system. The model's specificity, accuracy, precision, and sensitivity metrics crossed 92%, 86%, 82%, and 76%, respectively. ’Almost perfect agreement’ with Fleiss kappa of 0.847 was obtained between automated and manual gradings. This pixel count-based grading of KOH images offers a novel, cost-effective solution for quantifying fungal load.
UR - http://www.scopus.com/inward/record.url?scp=85207665246&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85207665246&partnerID=8YFLogxK
U2 - 10.1016/j.diagmicrobio.2024.116565
DO - 10.1016/j.diagmicrobio.2024.116565
M3 - Article
AN - SCOPUS:85207665246
SN - 0732-8893
VL - 111
JO - Diagnostic Microbiology and Infectious Disease
JF - Diagnostic Microbiology and Infectious Disease
IS - 1
M1 - 116565
ER -