Missingness as Signal: Channel-Independent Spectrogram Learning for Clinical Time Series Prediction
2026-07-03 • Machine Learning
Machine Learning
AI summaryⓘ
The authors developed a new method called CISM to predict patient outcomes in intensive care by analyzing time series data from different clinical measurements. Instead of just looking at the data values, they also use the pattern of which measurements are missing as important clues. Their method turns each variable into a special image showing how it changes over time and frequency, and it includes a stream to track missing data explicitly. Tests showed that CISM outperforms other methods, and the authors found that including missing data patterns helps make better predictions.
clinical time seriesmissing dataintensive care unittime-frequency spectrogramAUROCAUPRCMIMIC-IVmultivariate predictionmask injection
Authors
Soyeon Park, Charmgil Hong
Abstract
Clinical time series prediction in intensive care units remains challenging due to heterogeneous physiological variables and informative missingness. The presence or absence of a measurement can reflect clinical decisions and patient severity, and thus missingness can serve as a predictive signal rather than a simple data artifact. This work presents CISM, a Channel-Independent Spectrogram framework with a Missingness stream for clinical multivariate time series prediction. CISM converts each clinical variable into a variable-wise time-frequency spectrogram, preserves variable identity through variable-aligned encoding, and aligns an explicit missingness stream with the spectrogram representation. Experiments on an in-hospital mortality task derived from MIMIC-IV show that CISM achieves the highest mean AUROC (0.7225), AUPRC (0.3308), and F1 (0.3808) among the compared time series, missingness-aware, vision, and time-frequency baselines. Ablation studies further show that observation patterns provide a meaningful informative signal. Pixel-level mask injection improves performance over plain spectrogram inputs and recovers much of this predictive value. The aligned missingness stream contributes a further, complementary gain in both AUROC and AUPRC. These results highlight the importance of modeling observation patterns as structured signals in clinical time series prediction.