Hybrid CNN-LSTM Framework for Breast Cancer Diagnosis: A Spatial-Sequential Analytical Approach

The detection of breast cancer proves essential because this cancer remains both common and lethal amongst patients. The current traditional methods which diagnose conditions work effectively but involve subjectivity along with lengthy procedures. This paper presents a novel hybrid system that uses Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks to perform breast cancer diagnosis through histopathological image analysis. .Xr. The CNN segment recognizes advanced spatial relationships found within tissue specimens while the LSTM component tracks sequential dependencies of spatial element features thus creating an improved system for detecting faint malignant characteristics. This research used the BreakHis dataset which contains 7,909 breast cancer images and benign and malignant class labels for evaluation of the proposed model. The implementation resulted in 93.2% accuracy together with 92.4% precision along with 91.5% recall and 91.9% F1-score and 94.3% AUC. Two individual models composed of standalone CNN and LSTM reached different accuracy rates where CNNS obtained 86.5% while LSTMs achieved 82.3%. With the combination of CNN and LSTM components the hybrid system succeeded in correctly detecting malignant samples at a rate of 92%. ROC curve analysis proved the CNN-LSTM model profitable due to its superior outcome where the CNN-LSTM curve maintained better positions than competition models for sensitivity combined with specificity metrics. The research shows that combining CNN and LSTM networks delivers effective breast cancer detection which becomes an automated clinical tool for medical staff to rely on.