Machine Learning Supports Breast Cancer Diagnosis Predictions

Michigan Technology University researchers developed a machine learning model that uses probability to classify breast cancer more accurately in histopathology images and evaluate the uncertainty of its predictions.

Breast cancer is the most common cancer with the highest mortality rate. Early detection and diagnosis can diminish the impact of the disease. However, classifying breast cancer using histopathology images is difficult due to bias in data and the unavailability of annotated data in large quantities.

Automatic detection of breast cancer using a convolutional neural network (CNN) has shown promise, but it is linked with a high risk of false positives and negatives.

Without a measure of confidence, false predictions from CNN could lead to poor outcomes. However, the new machine learning model developed by Michigan Technology University researchers can evaluate the uncertainty in its predictions to classify benign and malignant tumors, reducing risk.

“Any machine learning algorithm that has been developed so far will have some uncertainty in its prediction,” mechanical engineering graduate student and study author, Ponkrshnan Thiagarajan, said in a press release.

“There is little way to quantify those uncertainties. Even if an algorithm tells us a person has cancer, we do not know the level of confidence in that prediction.”

According to researchers, not knowing how confident an algorithm is has made it difficult to rely on computer-generated predictions. The present model is an extension of the Bayesian neural network that can evaluate an image and produce an output.

The Michigan Tech model differentiates between negative and positive classes by examining the images, which at their most basic level are groups of pixels. Additionally, the model can measure the uncertainty in its predictions.

“Breast cancer is one of the cancers that has the highest mortality and highest incidence,” Thiagarajan said. “We believe that this is an exciting problem wherein better algorithms can make an impact on people’s lives directly.”

The researchers plan to expand the model for multiclass classification of breast cancer moving forward. Their goal is to detect cancer subtypes in addition to classifying benign and malignant tissue. The machine learning model could also be developed for other medical diagnoses.

“Despite the promise of machine learning-based classification models, their predictions suffer from uncertainties due to the inherent randomness and the bias in the data and the scarcity of large datasets,” assistant professor of mechanical engineering Susanta Ghosh said. “Our work attempts to address these issues and quantifies, uses, and explains the uncertainty.”