Multiple-Time-Series Clinical Data Processing for Classification With Merging Algorithm and Statistical Measures.

Area of Science:

• Biomedical Informatics
• Machine Learning in Medicine
• Oncology Data Analysis

Background:

• Traditional data processing in patient conditions can obscure clinical information, reducing prediction accuracy.
• Accurate prediction of hepatocellular carcinoma (HCC) recurrence is crucial for effective patient management.
• Existing methods may not fully leverage the value of multiple, time-series clinical measurements.

Purpose of the Study:

• To introduce a novel multiple-time-series data-processing algorithm with period merging.
• To enhance the accuracy of clinical-outcome prediction, specifically for HCC recurrence.
• To evaluate the effectiveness of the proposed algorithm in improving prediction performance.

Main Methods:

• A new multiple-time-series data-processing algorithm with period merging was developed.
• Clinical data from 83 HCC patients were processed using the merging algorithm.
• Multiple measurements support vector machine (MMSVM) with RBF kernels and multiple measurements random forest regression (MMRF) were employed for prediction.

Main Results:

• The proposed merging algorithm significantly improved HCC recurrence prediction accuracy.
• MMSVM with RBF using multiple measurements (120-day period) achieved optimal prediction (accuracy 0.771, balanced accuracy 0.603).
• This performance was statistically superior to predictions using single measurements (accuracy 0.626, balanced accuracy 0.459, P < 0.01).

Conclusions:

• The developed data-processing algorithm significantly enhances HCC recurrence prediction performance.
• Utilizing multiple measurements, processed via the novel algorithm, offers greater value than single measurements.
• This approach holds promise for improving clinical-outcome prediction in oncology.