Autonomous Decision Systems For Healthcare Diagnostics: Balancing Accuracy And Computational Efficiency

Abstract

An Autonomous Decision System (ADS) has gained increased application in the medical field as a means of improving early detection of diseases, but most deep learning architectures are computationally expensive and have low efficiency when used in a real-world medical context. The dataset employed is that from Kaggle, which provides Computed Tomography (CT) scans of lung cancer with several classes of images, such as normal, benign, and malignant states. In light of the stated problem, the research introduces an efficient Siberian Tiger Optimization–based Intelligent Lightweight Convolutional Neural Network (ST-IntLCNN) model for detecting  cancer in lung using CT images through feature extraction and classification. This model combines Contrast Limited Adaptive Histogram Equalization (CLAHE) is used for image handling stage and Histogram of Oriented Gradients (HOG) is used for dimensionality reduction. The features extracted from the images are classified using Intelligent Lightweight Convolutional Neural Networks (Int LCNNs).The design of the IntLCNN model with a lightweight architecture helps minimize memory consumption without compromising its ability to learn features effectively. Experimental results confirm that the developed ST-IntLCNN model provides excellent classification results with 98% accuracy and 97.98% precision, recall, and F1 score. The implementation is done using Python with the help of TensorFlow, Keras, OpenCV, and Scikit-learn for data preprocessing and model creation. The proposed ST-IntLCNN is capable of striking an ideal balance between performance and speed when classifying lung cancer from CT scans, thereby enabling its application in self-sufficient healthcare systems without the shortcomings of existing deep learning models.