Skin Cancer Segmentation and Classification Using Vision Transformer for Automatic Analysis in Dermatoscopy-Based Noninvasive Digital System

Algorithm 1

Step by step skin cancer segmentation and classification using ViT.

1. Input: I_DERMATOSCOPY, SegNet_SAM

I_DERMATOSCOPY= the 20X Dermatoscopy images

SegNet_SAM= the SAM segmentation

2. Initialization:

 i. I = Resize I_DERMATOSCOPYinto 224 × 224-pixel

 ii. I_SEG = Apply SegNet_SAM on I

 iii. I_SEG = Apply Binary Mask on I_SEG

3. ViT:

 i. Input: I_SEG

 ii. I_PATCH = Convert I_SEG into 14×14 patches

 iii. While m: = I_PATCH ≤ I_PATCH Count

  i. I_LP[m] = I_{PATCHUNROLLED}[m] × E [][]

  ii. D = Concatenate(E_POS[m], I_LP[m])

 iv. End While

 v. While n <= L do

  i. D'_n = MSA (LN(D_n-1)) + D_n-1

  ii. D_n = MLP (LN (D'_n)) + D'_n

 vi. End While

 vii. I_FINAL = LN(D_L)

4. If I_FINAL≤ Threshold Value Then

   Result: = “Benign”

Else

   Result: = “Malignant”

End If

5. Return Result