"Young Colorectal Cancer Patient Risk of Death, Predicting With Quantum Computing Technology"

Jul 12, 2024

(Figure left-hand table) Analysis of the prediction accuracy shows that the quantum machine learning model (QSVM (90%) has higher accuracy in predicting the risk of death in patients with early-onset colorectal cancer than the conventional machine learning model (Conventional SVM (70%)). It was confirmed that the prediction accuracy of quantum machine learning was maintained even though the outcome variables became sparse (Figure right-hand table).

A quantum machine learning model has been developed that can predict the risk of death in early-onset colorectal cancer patients.

Professor Park Yoo-rang and Dr. Yoo Jae-yong, researcher Shim Woo-seop and Professor Kim Han-sang of Yonsei Cancer Hospital announced on the 12th that they have developed a quantum machine learning model that predicts the risk of death based on clinical data of early-onset colorectal cancer patients, and the prediction accuracy reaches 90%.

Also called young colorectal cancer, 'early onset colorectal cancer' refers to colorectal cancer that occurs under the age of 50. The incidence of colorectal cancer in Korea's 20s and 40s is 12.9 per 100,000 people, ranking first in the world. Early-onset colorectal cancer is more aggressive and has a lower survival rate compared to colorectal cancer diagnosed in other age groups. Therefore, it is important to detect diseases early and treat them through accurate prognosis prediction.

Recently, various AI-based artificial intelligence models that can be used for disease diagnosis and prognosis prediction have been developed in the healthcare field. Sufficient clinical data is essential to increase the prediction accuracy of artificial intelligence models. However, the healthcare sector has difficulties such as cost problems and lack of data on rare diseases. For this reason, quantum computing technology that can increase analysis accuracy even with a small amount of data is attracting attention.

The research team developed a mortality risk prediction model for early-onset colorectal cancer patients through the quantum computing-based 'Quantum Support vector machine' and analyzed its accuracy.

Based on treatment data from 1,253 patients with early-onset colorectal cancer who visited Severance Hospital from 2008 to 2020, we developed a quantum machine learning model to predict the risk of death according to patients' disease conditions. A total of 93 variables, including patient information data such as age and gender, and clinical data on staging and treatment information, were applied as predictors of machine learning models using quantum computing technology.

To confirm the effectiveness of the model, the research team compared and analyzed the accuracy according to the optimal number of variables, sample size, and ratio of outcome variables with existing machine learning models. Prediction accuracy was analyzed as an indicator of the 'receiver operating characteristic curve' (AUROC). AUROC is a method of representing the predictive accuracy of a specific test tool for predicting any prognosis, with 'the area under the ROC curve'. It is mainly used as a performance evaluation index for AI models, and generally, the closer it is to 1, the better the performance, and if it is 0.8 or higher, it is evaluated as a high-performance model.

As a result of the analysis, the prediction accuracy of conventional machine learning models (Conventional SVMs) recorded 70%, while quantum machine learning models recorded 90% accuracy in predicting mortality risk for early-onset colorectal cancer patients.

In addition, the research team conducted performance verification by adjusting the ratio of death and survival to verify the robustness of quantum computing.

As a result, existing machine learning models showed 80% predictive performance when disproportionately adjusted the mortality rate. On the other hand, we found that the prediction accuracy of quantum machine learning models maintains a high prediction accuracy of 88% even when the death rate is imbalanced, and that quantum machine learning models maintain higher prediction accuracy compared to traditional machine learning models even when the death-survival rate is imbalanced.

Professor Yoo-rang Park "This study established a quantum machine learning model that accurately predicts the risk of death in patients with early-onset colorectal cancer."Based on this, we expect that quantum machine running models will continue to be used to expand into various areas of healthcare in the future "

Professor Kim Han-sang said, `This study is an example of digital healthcare using quantum computers and medical artificial intelligence in the field of tumors, and the introduction of digital healthcare technology for cancer diagnosis, treatment, and survivor management could change the treatment paradigm of cancer treatment sites in the future.'

On the other hand, this study was conducted with the support of 2024 research funds from the Ministry of Trade, Industry and Energy's Industrial Innovation Human Resources Growth Support Project (P0023675), and the research results were recently published in the international academic journal 『Applied Soft Computing』 (IF 8.7).