We propose Augmented Intelligence Metrics (AI Metrics) to interpret the output of CNN results.  The output and index of the CNN results were defined and calculated using the following methods:

malignancy output = a sum of model outputs for 5 malignant tumorous disorders (melanoma, BCC, SCC, intraepithelial carcinoma, and keratoacanthoma)
benign output = a sum of model outputs for 40 benign tumorous disorders (seborrheic keratosis, wart, etc.)
steroid output = a sum of model outputs for 59 disorders requiring a steroid treatment (atopic dermatitis, contact dermatitis, etc.)
antibiotics output = a sum of model outputs for 21 disorders requiring an antibiotic treatment (cellulitis, impetigo, etc.)
antiviral output = a sum of model outputs for 5 disorders requiring an antiviral treatment (herpes simplex, herpes zoster, etc.)
antifungal output = a sum of model outputs for 8 disorders requiring an antifungal treatment (tinea pedis, tinea versicolor, etc.)

malignancy index = malignancy output / (malignancy output + benign tumorous disorders output)
steroid index = steroid output / (steroid output + antibiotics output + antifungal output + antiviral output)
antibiotics index = antibiotics output / (steroid output + antibiotics output + antifungal output + antiviral output)
antiviral index = antiviral output / (steroid output + antibiotics output + antifungal output + antiviral output)
antifungal index = antifungal output / (steroid output + antibiotics output + antifungal output + antiviral output)

Leave a Reply