Precise separation of areas in medical imaging is crucial in the field of medical operations, such as surgical planning, diagnostics, condition, degenerative disc disease, kyphosis, and spinal curvatures. It is also important in assessing patients following surgery. Despite the high contrast of bone structures in medical pictures, segmenting patches can be difficult because of their complicated structure, irregular spine curvature, and ill-defined borders. Deep learning (DL) has been used extensively in the past several years for segmenting pictures of patches. This work has introduced a model for automated vertebral HAM10000 image segmentation using Patch Classification Fine Tuned Inception V4 (PC-FT Inception V4) and overlapping patches, which represents a major step towards a robust and automated approach. This work has used overlapping patches in segmentation tasks using Patch PC-FT Inception Net because 3D convolutional neural networks have higher memory and processing costs and are more prone to overfitting. The local information included in HAM1000 images is efficiently preserved by PC-FT Inception Net in the proposed patch segmentation approach. We employ the Random Sampling Method to equally distribute the classes over the square patches created by dividing the overlapping HAM picture slices in order to reduce the amount of processing resources required. These patches are then added to the model together with the matching ground truth patches. The HAM10000 dataset challenge has evaluated this method using publicly available HAM images. The results indicate that PC-FT Inception Net has a precision of 92.1%, septicity of 99%, accuracy of 99.82%, F1-Score of 95.1% in terms of the patch-based classification accuracy, and Jaccard index (JAC) of 94.32%, Dice Similarity Score (DSC) of 95%. The IoU of 99%, and mIoU of 97.9% in terms of the division precision that outflank past strategies over all measurements.