Deep Hybrid Models in Medical Imaging: CNN-SVM-Based Lung Cancer Detection with Statistical Insights

Lung cancer stands as one of the leading deadly malignancies across the globe which demonstrates why early identification through accurate diagnosis procedures is so important. The authors developed a learning model which combines Convolutional Neural Networks (CNN) and Support Vector Machines (SVM) to detect lung cancer in CT scan images more effectively. The CNN-SVM model performs better than individual models because it leverages SVM classification capabilities with CNN feature extraction abilities. The devised system received evaluation on an existing database of 8000 CT scan images which contained four distinct groups: benign, malignant, metastatic, and normal. The combination performance of CNN-SVM overstands individual models to yield a 92.3% accuracy rate with precision at 91.5% and recall at 91.7% and F1-score reaching 91.6%. The standalone CNN model reached 85.4% accuracy but the SVM model produced 78.6% accuracy. The hybrid CNN-SVM model demonstrated superior performance at all thresholds. Paired t-tests showing improvements revealed that the differences between CNN and CNN-SVM resulted in a t-statistic of -2.916 (p-value = 0.0036) and the SVM and CNN-SVM differences yielded -5.559 (p-value =3.46×10−8). oord-CAM visualization tools demonstrate how the model effectively detects essential tumor zones which enhances its clinical significance as well as its interpretability potential. The combination model of CNN-SVM demonstrates a strong and efficient approach for lung cancer detection which can find practical use in clinical diagnosis.