Hospital Length of Stay Prediction Based on
Multi-modal Data Towards Trustworthy Human-AI Collaboration in Radiomics


 Hubert Baniecki\(^{1,2}\), Bartlomiej Sobieski\(^{2}\), Przemysław Bombiński\(^{2,3}\), Patryk Szatkowski\(^{2,3}\), Przemysław Biecek\(^{1,2}\)


\(^{1}\)University of Warsaw, \(^{2}\)Warsaw University of Technology,
\(^{3}\)Medical University of Warsaw


 AIME 2023, Portoroz, Slovenia

June 13, 2023

  • PhD student in computer science at the University of Warsaw, Poland
  • research interest: explainable machine learning
  1. How valuable are features extracted from medical images?
  2. How to explain machine learning survival models?

Predicting hospital length of stay (LoS)

To what extent can the patient’s length of stay in a hospital be predicted using only an X-ray image?


Motivation: explain black-box models predicting hospital LoS

K. Stone et al. PLOS Digital Health, 1(4):e0000017, 2022.


Task: time-to-event prediction

instead of a single-value time regression or time-span classification

Dataset i.e. tabular data openly available on GitHub

We manually annotated textual radiology reports of X-ray images:

  • 1235 patients from one of the Polish hospitals
  • target feature: time between the patient’s radiological examination and hospital discharge (in days, \(min=1\), \(median=7\), \(mean=13.73\), \(max=330\)); about 20% of outcomes are right-censored
  • 95 features:
    • 2 baseline: age & sex
    • 17 human-annotated pathology occurrences, e.g. pulmonary nodule, pleural effusion, medical devices
    • 76 algorithm-extracted image statistics using the pyradiomics tool J. Van Griethuysen et al. Cancer Research, 77(21):e104–e107, 2017.

Benchmarking machine learning models

Predicting the patient’s LoS from an X-ray image is indeed possible, but challenging.

Multi-modal feature performance

Performance of the interpretable model declines when increasing the number of features.

How to explain hospital length of stay?


For classification (or regression), we have:


What about time-to-event prediction?

SurvSHAP(t)

M. Krzyziński, M. Spytek, H. Baniecki, P. Biecek. SurvSHAP(t): Time-dependent explanations of machine learning survival models. Knowledge-Based Systems, 262:110234, 2023.


Implemented in both R & Python!

Explaining LoS predictions to humans

Feature importance i.e. an alternative to p-values

Limitations and future work

  1. The segmentation model used to obtain masks for feature extraction propagates errors to LoS predictions, and it can be improved if the final goal of a study would be a definitive evaluation of those features.
  1. Tuning the hyperparameters of survival models is an option.
  1. One needs to be sure that humans properly interpret explanation visualizations, i.e. physicians and other stakeholders interpreting predictions need to understand that explanations are only an approximation of the black-box model.

Key takeaways

  1. X-ray images encode valuable information for predicting hospital LoS
  2. Interpret survival models using time-dependent explanations
  1. The TLOS dataset is openly available for research purposes github.com/mi2datalab/xlungs-trustworthy-los-prediction


Contact and resources: hbaniecki.com

References

  • Z. Huang et al. Length of stay prediction for clinical treatment process using temporal similarity. Expert Systems with Applications, 40(16):6330–6339, 2013.
  • K. Stone et al. A systematic review of the prediction of hospital length of stay: Towards a unified framework. PLOS Digital Health, 1(4):e0000017, 2022.
  • J. Van Griethuysen et al. Computational radiomics system to decode the radiographic phenotype. Cancer Research, 77(21):e104–e107, 2017.
  • M. Krzyziński, M. Spytek, H. Baniecki, P. Biecek. SurvSHAP(t): Time-dependent explanations of machine learning survival models. Knowledge-Based Systems, 262:110234, 2023.

Hubert Baniecki et al.   Hospital Length of Stay Prediction..   pp. 65–74