This study investigates the machine learning techniques for unsupervised image classification and quality assessment in the domain of ultrasound imaging. Leveraging Convolutional Neural Networks (CNNs) for feature extraction and subsequent integration into a Support Vector Machine (SVM) model, we explored a novel approach aimed at accurate image classification. The dataset comprises high-frequency images in the form of image sequences depicting the facial skin of females. The study's primary emphasis was to categorize ultrasound images based on learned deep features, offering a distinctive framework for unsupervised image classification. The investigation employed CNNs to extract deep features from images, enhancing the SVM model's performance in accurately categorizing images. The incorporation of gamma correction as a preprocessing step further augmented the accuracy and sensitivity of the models. The SVM model exhibited exceptional performance, achieving accuracy rates exceeding 95.43% in the training phase and approximately 94.72% during testing, that is a significant milestone in the precise classification of ultrasound images.