Is there a public disease mapping instructions?

[nina-weng]

Thanks for the developing of this repository! I noticed that the download meta data consists around 20-ish diagnosis, I was wondering whether there is a mapping instructions which could categorize those into the 9 categories in ISIC challenge? i.e.,

1. Melanoma
2. Melanocytic nevus
3. Basal cell carcinoma
4. Actinic keratosis
5. Benign keratosis (solar lentigo / seborrheic keratosis / lichen planus-like keratosis)
6. Dermatofibroma
7. Vascular lesion
8. Squamous cell carcinoma
9. None of the others

Any information will be appreciated!

[nina-weng]

I tried to map by myself, however what the internet told me is that other than the disease mentioned in the 9 categories above, the others are just not included Not sure if that's the real case in Dermatology

[NickRaymondKurtansky]

@nina-weng Some diagnoses outside of 1:8 were used to make up the out-of-distribution class (9 None of the others) and some diagnoses were not included at all for the ISIC 2019 Challenge. I'm not sure what you mean by, "Not sure that's the real case in Dermatology," but when making a prediction about a diagnoses or whether or not the lesion is harmful, a skilled clinician usually narrows it down to two or three possible diagnoses that are visually similar. In theory, therefore, the skilled physician would not need an AI that classifies broadly over every potential skin condition. The motivation for selecting those 8 classes and 1 out-of-distribution class is because it contains the three main malignant skin neoplasms (melanoma, basal cell, and squamous cell carcinoma) and five other diagnoses that are very often considered in the clinical differential for one or more of those three malignancies.

[nina-weng]

Thank you so much for this information! When I said "not sure whether that's the real case in dermatology", I was referring to the fact that when I searched, e.g. which type of the following categories that 'AIMP' belongs to, it told me it does not belong to any of the above 8 categories. I was just wondering whether that's correct from the dermatology's perspective. Now I understand the situation here. thanks a lot

[NickRaymondKurtansky]

Atypical intraepidermal melanocytic proliferation (AIMP) is a descriptive histopathologic term commonly used in cases to denote morphology sharing some features with melanoma but failing to meet criteria of a definitive benign or malignant diagnosis (Ensslin et al. Atypical Melanocytic Proliferations: A Review of the Literature. Dermatol Surg. 2018).

Some people consider it an intermediate diagnosis while some people are reluctant to use language implying progression since the malignant potential of these lesions is ambiguous.