Abstract
Medical images of brain tumors are critical for characterizing the pathology of tumors and early diagnosis. There are multiple modalities for medical images of brain tumors. Fusing the unique features of each modality of the magnetic resonance imaging (MRI) scans can accurately determine the nature of brain tumors. The current genetic analysis approach is time-consuming and requires surgical extraction of brain tissue samples. Accurate classification of multi-modal brain tumor images can speed up the detection process and alleviate patient suffering. Medical image fusion refers to effectively merging the significant information of multiple source images of the same tissue into one image, which will carry abundant information for diagnosis. This paper proposes a novel attentive deep-learning-based classification model that integrates multi-modal feature aggregation, lite attention mechanism, separable embedding, and modal-wise shortcuts for performance improvement. We evaluate our model on the RSNA-MICCAI dataset, a scenario-specific medical image dataset, and demonstrate that the proposed method outperforms the state-of-the-art (SOTA) by around 3%.
Highlights
GLOBOCAN recently conducted a survey in 185 countries, reporting an estimation of over 300 K new brain cancer cases and above 250 K new deaths in 2020 [1]
We propose an attentive multi-modal deep neural network (DNN) to predict the status of the methylguanine DNA methyltransferase (MGMT) promoter methylation
We focus on the RSNA-MICCAI dataset [9], a multi-center brain tumor magnetic resonance imaging (MRI) dataset that comes with two tasks; namely, tumor segmentation and MGMT detection
Summary
GLOBOCAN recently conducted a survey in 185 countries, reporting an estimation of over 300 K new brain cancer cases and above 250 K new deaths in 2020 [1]. In addition to a multi-modal feature aggregation strategy, our proposed model integrates three performance boosters, including a lite attention mechanism to control the model size and speed up training, a separable embedding module to improve the feature representation of MRI data, and a modal-wise shortcut strategy to ensure the modal specificity. These joint efforts have improved the detection accuracy of our model by 3%, compared to the SOTA method.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have