With the availability of big image datasets and state-of-the-art computing hardware, data-driven machine learning approaches, particularly deep learning, have been used in numerous medical image (CT-scans, MRI, PET, SPECT, etc..) computing tasks, ranging from image reconstruction, super-resolution, segmentation, registration all the way to disease classification and survival prediction. However, training such high-precision approaches often require large amounts of data to be collected and labelled and high-capacity graphics processing units (GPUs) installed, which are resource intensive and hence not always practical. Other hurdles such as the generalization ability to unseen new data and difficulty to interpret and explain can prevent their deployment to those clinical applications which deem such abilities imperative.
Over the past decades or so, model-driven approaches that utilize prior knowledge have been extensively studied and developed for medical image computing and analysis. Recent advances in machine learning (particularly deep learning) however have shifted the tendency and led to a number of breakthroughs in high-level computer vision tasks that have been previously considered too difficult to tackle. By leveraging unprecedented big data, machine learning approaches have repeatedly proven more accurate and efficient than their model-driven counterparts. On the other hand, approaches driven by prior knowledge often offer other advantages over data-driven machine learning approaches, including better generalization, clear interpretation and explainability, and strong robustness. As such, the combination of data-driven approaches with prior knowledge is likely to overcome the problems of data-driven approaches and therefore lead to further advances in medical image computing and analysis.
The expected topics may include, but are not limited to the following:
• One-short and few-short learning
• Image registration and segmentation
• Image acquisition and reconstruction
• Interpretable, explainable, and causal learning
• Statistical and computational geometric modelling
• Uncertainty modelling and quantification
• Novel data-driven or model-driven methods for medical imaging
• Inference and learning under uncertain, incomplete or limited data
• Representation learning, image synthesis, and generative modelling
With the availability of big image datasets and state-of-the-art computing hardware, data-driven machine learning approaches, particularly deep learning, have been used in numerous medical image (CT-scans, MRI, PET, SPECT, etc..) computing tasks, ranging from image reconstruction, super-resolution, segmentation, registration all the way to disease classification and survival prediction. However, training such high-precision approaches often require large amounts of data to be collected and labelled and high-capacity graphics processing units (GPUs) installed, which are resource intensive and hence not always practical. Other hurdles such as the generalization ability to unseen new data and difficulty to interpret and explain can prevent their deployment to those clinical applications which deem such abilities imperative.
Over the past decades or so, model-driven approaches that utilize prior knowledge have been extensively studied and developed for medical image computing and analysis. Recent advances in machine learning (particularly deep learning) however have shifted the tendency and led to a number of breakthroughs in high-level computer vision tasks that have been previously considered too difficult to tackle. By leveraging unprecedented big data, machine learning approaches have repeatedly proven more accurate and efficient than their model-driven counterparts. On the other hand, approaches driven by prior knowledge often offer other advantages over data-driven machine learning approaches, including better generalization, clear interpretation and explainability, and strong robustness. As such, the combination of data-driven approaches with prior knowledge is likely to overcome the problems of data-driven approaches and therefore lead to further advances in medical image computing and analysis.
The expected topics may include, but are not limited to the following:
• One-short and few-short learning
• Image registration and segmentation
• Image acquisition and reconstruction
• Interpretable, explainable, and causal learning
• Statistical and computational geometric modelling
• Uncertainty modelling and quantification
• Novel data-driven or model-driven methods for medical imaging
• Inference and learning under uncertain, incomplete or limited data
• Representation learning, image synthesis, and generative modelling