When will the mist clear? On the Interpretability of Machine Learning for Medical Applications: a survey (2010.00353v1)

Abstract: Artificial Intelligence is providing astonishing results, with medicine being one of its favourite playgrounds. In a few decades, computers may be capable of formulating diagnoses and choosing the correct treatment, while robots may perform surgical operations, and conversational agents could interact with patients as virtual coaches. Machine Learning and, in particular, Deep Neural Networks are behind this revolution. In this scenario, important decisions will be controlled by standalone machines that have learned predictive models from provided data. Among the most challenging targets of interest in medicine are cancer diagnosis and therapies but, to start this revolution, software tools need to be adapted to cover the new requirements. In this sense, learning tools are becoming a commodity in Python and Matlab libraries, just to name two, but to exploit all their possibilities, it is essential to fully understand how models are interpreted and which models are more interpretable than others. In this survey, we analyse current machine learning models, frameworks, databases and other related tools as applied to medicine - specifically, to cancer research - and we discuss their interpretability, performance and the necessary input data. From the evidence available, ANN, LR and SVM have been observed to be the preferred models. Besides, CNNs, supported by the rapid development of GPUs and tensor-oriented programming libraries, are gaining in importance. However, the interpretability of results by doctors is rarely considered which is a factor that needs to be improved. We therefore consider this study to be a timely contribution to the issue.

• The paper identifies critical challenges in interpreting ML models for cancer diagnosis and emphasizes the need for transparent decision frameworks.
• It reviews various ML approaches, noting CNN's effectiveness with an AUC of 0.81 in skin cancer risk prediction.
• The study advocates for using explainable AI and feature selection to enhance clinical trust and the integration of ML in healthcare.

Interpretability in Machine Learning for Medical Applications: A Survey

The survey paper titled "When will the mist clear? On the Interpretability of Machine Learning for Medical Applications" addresses the increasingly critical issue of interpretability in the deployment of ML models in medical applications, with a focused discussion on cancer research. The paper emphasizes the importance of interpretability in the context of complex decision-making processes in medicine, where ML systems must align with clinical practices and gain the trust of medical professionals.

The paper provides a comprehensive analysis of various ML frameworks, models, databases, and tools applied to medical applications, highlighting interpretability as a key factor often overlooked in current implementations. It reviews a broad spectrum of ML models, including Artificial Neural Networks (ANN), Logistic Regression (LR), Support Vector Machine (SVM), and Convolutional Neural Networks (CNN), particularly as they pertain to cancer diagnosis and treatment prediction.

Key Numerical Findings

Among the models examined, Artificial Neural Networks (ANN), Logistic Regression (LR), and Support Vector Machines (SVM) emerge as frequently utilized in cancer prediction scenarios. However, the paper highlights CNNs' growing importance due to the rapid development of GPUs and tensor-based programming libraries, which facilitate advanced image processing crucial in medical diagnostics. For instance, CNNs have reported Area Under the Curve (AUC) values of 0.81 in skin cancer risk prediction, emphasizing their effectiveness.

Interpretability Challenges and Proposals

The paper identifies several challenges related to the interpretability of ML models in the medical domain:

1. Output Explainability: There is a growing demand for machine learning outputs to be interpretable by medical professionals. Despite the development of sophisticated models, their adoption in clinical settings is conditional on the clarity of the results provided.
2. Linking Outputs to Inputs: The ability of models to explain how specific inputs contribute to a given output is crucial. This connection helps clinicians understand decision-making processes and trust the AI's recommendations.
3. Data Hungriness: While deep learning models rely heavily on large data volumes, the medical field often deals with limited datasets due to privacy concerns and data acquisition challenges.

The paper suggests that enhanced model interpretability and incorporating methods like feature selection before model deployment could aid in addressing these issues. Additionally, XAI (Explainable AI) strategies offer potential pathways to alleviate the interpretability gap, making complex models more transparent.

Implications and Future Directions

The survey argues for a directed effort toward improving ML model interpretability to ensure integration into clinical practice. This involves developing models that balance performance with explainability, ensuring ML systems contribute meaningfully to medical decision-making. The research underscores that while ANN, LR, and SVM remain prevalent, there is significant scope for CNNs, particularly in imaging applications due to their proficiency in handling complex image data.

Future work is suggested to focus on enhancing computational frameworks, such as GPUs, to improve algorithmic speed and efficiency. Moreover, the development of explainable methodologies in ML could democratize AI technology, extending its utility across various domains of medical research and practice. The paper’s review suggests that overcoming interpretability barriers is key to unlocking the full potential of machine learning in personalized medicine.

In conclusion, the paper offers valuable insights into the current state and future landscape of machine learning applications in medicine, emphasizing the pivotal role of interpretability in widespread adoption within clinical settings. The integration of high-performance computing and improved transparency can indeed transform the application of ML in medical sciences, making predictive and diagnostic tools both powerful and trusted allies in healthcare.