Abstract:Cancer is a significant health concern globally and its early identification and screening are essential for better patient outcomes. To accurately perform cancer diagnosis and ascertain prognostic as well as therapeutic targets, histopathology images are used. Deep learning techniques have had significant success which has spurred substantial research into, among other cancer illnesses, breast, colon, and lung cancer, but their effectiveness is limited by drawbacks such as the need to select appropriate features, which can be time-consuming and challenging. Furthermore, despite the success of deep learning models, their decisions are opaque. There is a need to understand their decision-making processes. This paper proposes a novel deep-learning model for histopathological medical image classification, specifically breast, colon, and lung cancer analysis based on the various input image characteristics (low-level and high-level features). The BreakHis dataset (Breast Cancer) and LC25000 (Colon and Lung Cancer) are used to evaluate the effectiveness of the proposed model. The proposed model's average accuracy, sensitivity, specificity, F1_score, recall, and precision are 98.00 +/- 0.19 with a specific accuracy of 99.2 % and 99.1 % (BreakHis 4 and 8 classes), 99.9 % and 100 % (LC25000 3 and 2 classes) respectively. In other to interpret the proposed model results, an explainable AI model based on LIME and SHAP is designed for a human-understandable explanation. Furthermore, parameter sensitivity analysis of the proposed model was recorded for both datasets. Results show that the proffered model supersedes the traditional methods that rely only on high-level features with a wide and significant margin as well as highlights its importance for accurate cancer analysis and aiding healthcare practitioners to identify and diagnose cancer at an early stage.

Explainability in CNN based Deep Learning models for medical image classification

Explainable Artificial Intelligence for Human Decision-Support System in Medical Domain

Achievements and Challenges in Explaining Deep Learning based Computer-Aided Diagnosis Systems

Explainable Artificial Intelligence (XAI) for Deep Learning Based Medical Imaging Classification

Explainability of deep learning models in medical video analysis: a survey

Is Grad-CAM Explainable in Medical Images?

Analysis of Explainable Artificial Intelligence Methods on Medical Image Classification

Are Explanations Helpful? A Comparative Analysis of Explainability Methods in Skin Lesion Classifiers

Exploring Explainable AI Techniques for Improved Interpretability in Lung and Colon Cancer Classification

Transparent and Accurate COVID-19 Diagnosis: Integrating Explainable AI with Advanced Deep Learning in CT Imaging

Explainability of Deep Neural Networks for Brain Tumor Detection

Enhancing Deep Learning Model Explainability in Brain Tumor Datasets using Post-Heuristic Approaches

DLIME: A Deterministic Local Interpretable Model-Agnostic Explanations Approach for Computer-Aided Diagnosis Systems

Grad-CAM-Based Explainable Artificial Intelligence Related to Medical Text Processing

Quantifying Explainable AI Methods in Medical Diagnosis: A study in skin cancer

Explainable Deep Learning Methods in Medical Image Classification: A Survey

Saliency-driven explainable deep learning in medical imaging: bridging visual explainability and statistical quantitative analysis

Enhancing Histopathological Medical Image Classification for Early Cancer Diagnosis Using Deep Learning and Explainable AI - LIME & SHAP

Explainable AI for Retinoblastoma Diagnosis: Interpreting Deep Learning Models with LIME and SHAP

Visual Interpretable and Explainable Deep Learning Models for Brain Tumor MRI and COVID-19 Chest X-ray Images

Aligning Human Knowledge with Visual Concepts Towards Explainable Medical Image Classification