An official website of the United States government
Brain networks have attracted increasing attention due to the potential to better characterize brain dynamics and abnormalities in neurological and psychiatric conditions. Recent years have witnessed enormous successes in deep learning. Many AI algorithms, especially graph learning methods, have been proposed to analyze brain networks. An important issue for existing graph learning methods is that those models are not typically easy to interpret. In this study, we proposed an interpretable graph learning model for brain network regression analysis. We applied this new framework on the subjects from Human Connectome Project (HCP) for predicting multiple Adult Self-Report (ASR) scores. We also use one of the ASR scores as the example to demonstrate how to identify sex differences in the regression process using our model. In comparison with other state-of-the-art methods, our results clearly demonstrate the superiority of our new model in effectiveness, fairness, and transparency.
more »
« less
3D Brain Tumor Segmentation: Narrow UNet CNN
Reliable methods for tumor detection and brain abnormalities are crucial to help find diseases at early stages. Having accurate software that uses machine learning to identify abnormalities of the brain may prevent a disease progression if used on an MRI of the patient. In this paper, we develop a neural network to detect and highlight brain tumors present in MRI’s. Our model is designed to be more compact than typical CNNs to minimize prediction times while still maintaining prediction accuracy. The model uses a dice coefficient for the loss function as well as accuracy metric. We adopt Adadelta as the optimizer as it is more robust and eliminates the requirement of manually tuned learning rates. Our model reduces the prediction time with fewer layers and convolution filters, while allowing rapid convergence to a stable solution. In addition, the models hyper-parameters are being fine-tuned in an iterative process to ideally achieves better segmentation accuracy. Experimental results show that our model improves the performance compared with the state-of-the-art methods.
more »
« less
- Award ID(s):
- 1757787
- PAR ID:
- 10095110
- Date Published:
- Journal Name:
- IEEE MIT URTC (Undergraduate Research Technology Conference)
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract Brain tumor is a life-threatening disease and causes about 0.25 million deaths worldwide in 2020. Magnetic Resonance Imaging (MRI) is frequently used for diagnosing brain tumors. In medically underdeveloped regions, physicians who can accurately diagnose and assess the severity of brain tumors from MRI are highly lacking. Deep learning methods have been developed to assist physicians in detecting brain tumors from MRI and determining their subtypes. In existing methods, neural architectures are manually designed by human experts, which is time-consuming and labor-intensive. To address this problem, we propose to automatically search for high-performance neural architectures for classifying brain tumors from MRIs, by leveraging a Learning-by-Self-Explanation (LeaSE) architecture search method. LeaSE consists of an explainer model and an audience model. The explainer aims at searching for a highly performant architecture by encouraging the architecture to generate high-fidelity explanations of prediction outcomes, where explanations’ fidelity is evaluated by the audience model. LeaSE is formulated as a four-level optimization problem involving a sequence of four learning stages which are conducted end-to-end. We apply LeaSE for MRI-based brain tumor classification, including four classes: glioma, meningioma, pituitary tumor, and healthy, on a dataset containing 3264 MRI images. Results show that our method can search for neural architectures that achieve better classification accuracy than manually designed deep neural networks while having fewer model parameters. For example, our method achieves a test accuracy of 90.6% and an AUC of 95.6% with 3.75M parameters while the accuracy and AUC of a human-designed network—ResNet101—is 84.5% and 90.1% respectively with 42.56M parameters. In addition, our method outperforms state-of-the-art neural architecture search methods.more » « less
-
Abstract Large language models (LLMs) have been shown to have significant potential in few-shot learning across various fields, even with minimal training data. However, their ability to generalize to unseen tasks in more complex fields, such as biology and medicine has yet to be fully evaluated. LLMs can offer a promising alternative approach for biological inference, particularly in cases where structured data and sample size are limited, by extracting prior knowledge from text corpora. Here we report our proposed few-shot learning approach, which uses LLMs to predict the synergy of drug pairs in rare tissues that lack structured data and features. Our experiments, which involved seven rare tissues from different cancer types, demonstrate that the LLM-based prediction model achieves significant accuracy with very few or zero samples. Our proposed model, the CancerGPT (with ~ 124M parameters), is comparable to the larger fine-tuned GPT-3 model (with ~ 175B parameters). Our research contributes to tackling drug pair synergy prediction in rare tissues with limited data, and also advancing the use of LLMs for biological and medical inference tasks.more » « less
-
Large-scale channel prediction, i.e., estimation of the pathloss from geographical/morphological/building maps, is an essential component of wireless network planning. Ray tracing (RT)-based methods have been widely used for many years, but they require significant computational effort that may become prohibitive with the increased network densification and/or use of higher frequencies in B5G/6G systems. In this paper, we propose a data-driven, model-free pathloss map prediction (PMP) method, called PMNet. PMNet uses a supervised learning approach: it is trained on a limited amount of RT data and map data. Once trained, PMNet can predict pathloss over location with high accuracy (an RMSE level of 10−2 ) in a few milliseconds. We further extend PMNet by employing transfer learning (TL). TL allows PMNet to learn a new network scenario quickly ( ×5.6 faster training) and efficiently (using ×4.5 less data) by transferring knowledge from a pre-trained model, while retaining accuracy. Our results demonstrate that PMNet is a scalable and generalizable ML-based PMP method, showing its potential to be used in several network optimization applications.more » « less
-
Pathloss prediction is an essential component of wireless network planning. While ray tracing based methods have been successfully used for many years, they require significant computational effort that may become prohibitive with the increased network densification and/or use of higher frequencies in 5G/B5G (beyond 5G) systems. In this paper, we propose and evaluate a data-driven and model-free pathloss prediction method, dubbed PMNet. This method uses a supervised learning approach: training a neural network (NN) with a limited amount of ray tracing (or channel measurement) data and map data and then predicting the pathloss over location with no ray tracing data with a high level of accuracy. Our proposed pathloss map prediction-oriented NN architecture, which is empowered by state-of-the-art computer vision techniques, outperforms other architectures that have been previously proposed (e.g., UNet, RadioUNet) in terms of accuracy while showing generalization capability. Moreover, PMNet trained on a 4-fold smaller dataset surpasses the other baselines (trained on a 4-fold larger dataset), corroborating the potential of PMNet.1more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
