Weakly supervised multitask learning models to identify symptom onset time of unclear-onset intracerebral hemorrhage

Abstract

Background Approximately 1/3 of spontaneous intracerebral hemorrhage (sICH) patients did not know the onset time and were excluded from studies about time-dependent treatments for hyperacute sICH. Aims To help clinicians explore the benefit of time-dependent treatments for unclear-onset patients, we presented artificial intelligence models to identify onset time using non-contrast computed tomography (NCCT) based on weakly supervised multitask learning (WS-MTL) structure. Methods The patients with reliable symptom onset time (strong label) or repeat CT (weak label) were included and split into training set and test set (internal and external). The WS-MTL structure utilized strong and weak labels simultaneously to improve performance. The models included three binary classification models for classifying whether NCCT acquired within 6, 8 or 12 hours for different treatments measured by area under curve (AUC), and a regression model for determining the exact onset time measured by mean absolute error (MAE). The generalizability of models was also explored in comprehensive analysis. Results 4 004 patients with 10 780 NCCT scans were included. The performance of WS-MTL classification model showed high accuracy, and that of regression model was satisfactory in ≤ 6 hours subgroup. In comprehensive analysis, the WS-MTL showed better performance for larger hematomas and thinner scans. And the performance improved effectively as training amounts increasing and could be improved steadily through retraining. Conclusions The WS-MTL models showed good performance and generalizability. Considering the large number of unclear-onset sICH patients, it may be worth to integrate the WS-MTL model into clinical practice to identify the onset time.

Evaluation of QCNN-LSTM for Disability Forecasting in Multiple Sclerosis using Sequential Multisequence MRI

“…This has traditionally been modeled using classical machine learning (ML) architectures ranging from Recurrent Neural Networks (RNNs) to the most recent innovation, transformers. In the medical imaging space, however, there have been few studies to predict disease progression based upon temporal data from imaging modalities such as Computed Tomography (CT) (Jianbo, 2022;Kim, 2021;Lucas, 2018), Magnetic Resonance Imaging (MRI) (Wang, 2024, Yan, 2022, and functional MRI (fMRI) (Yoon, 2020).…”