Deep learning (DL) is assisting academicians and medical professionals in uncovering latent opportunities in data and enhancing the healthcare industry. The edge computing applications like smart healthcare systems where accurate decision-making is required for fast medical treatment. DL in healthcare allows clinicians to correctly analyze any ailment and treat it, resulting in improved medical decisions. We present a unique DL model for the autonomous healthcare edge computing application in this paper. Computer Aided Diagnosis (CAD) is an essential requirement of healthcare edge computing where the patient's medical data is used for fast and accurate disease prediction. Propose the DL-based CAD model for automatic disease classification from the input medical images. The model consists of pre-processing, DL-based feature engineering, and classification. Input medical image is first pre-processed for quality improvement and then automatic features are extracted using the pre-trained DL models (ResNet50 and Densenet201). The pre-trained models are improved by performing the feature scaling followed by a separate classification phase. The proposed CAD model is experimentally evaluated using the medical images dataset. The results reveal the efficiency of the proposed model compared to underlying solutions.