Hybrid Deep Learning Model for Prostate Cancer Detection Using LeNet and MobileNet V2 Algorithms

Cancer of the prostate is among the most common neoplasms worldwide and in men especially; therefore, early diagnosis enhances the chances of the patients. In this work, we propose a proposed deep-learning framework by integrating LeNet and MobileNet V2 architectures for early diagnosis of prostate cancer from medical images. The proposed model had an accuracy of 96.2% and precision of 97.9<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>, a recall of 96.9% and F1-Score of 97.4% and surpassed LeNet, MobileNet V2 and ResNet-50 and DenseNet-121 models. The particularity of the test was confirmed with a confusion matrix that revealed 15 false negatives and 10 false positives. From the computational perspective, the model has an inference time of roughly 48 ms for per image and has a memory consumption of 60 MB, it is thus fast for real time applications in clinics. It adds more explainability techniques such as Grad-CAM, which improves interpretability, which is essential for clinician trust. This work shows that PFCLA can be used in early diagnosis of prostate cancer using enhanced deep learning hybrid model and future works that can be done include, Exploratory Study of More real world independent data set on PFCLA along with, Expanding the applicability of PFCLA to other medical imaging tasks.