Check-in #3 Reflection
Introduction:
The detection of abdominal trauma from Computed Tomography (CT) is an important area in the application of AI in medical imaging. The accuracy of deep learning models in analysis of 3D images heavily depends on the quality and representation of the training data. Therefore, using the appropriate method to sample frames from volumetric CT scans becomes an important step in model training. Different frame sampling techniques are developed to process sequential data types like video, specifically uniform sampling, random sampling, and attention-based sampling. Uniform sampling selects frames at regular intervals throughout the series. Random sampling involves randomly selecting frames to introduce variability into the training data. Attention-based sampling uses attention mechanisms to dynamically select frames based on their relevance. In this project, we aim to investigate these three different frame sampling strategies for the detection of organ-level injuries in abdominal CT scans. Through analyzing and comparing the effects of different sampling techniques on model’s accuracy, robustness, and efficiency, we aim to uncover insights that can inform the development of more effective sampling approaches for abdominal trauma detection.
Challenges:
During the process of completing the final project, we encountered many challenges. For instance, we had to deal with a large amount of unlabeled organ data. We needed to locate each organ/organ labels based on the segmentation dataset within a massive database. Additionally, we designed different sampling strategies to ensure the model's accuracy. More importantly, we had to construct a training pipeline.This pipeline is not only involved modifying the model's structure and reshaping the data, but also included experimenting with different optimizer and activation functions. Therefore, we learn and progress through difficulties.
Insights:
Currently, we are only using loss as a reference to refine our model and pipeline. For now, we have completed 70% of the entire pipeline and we will attempt to insert LSTM and Attention/Mask blocks into our pipeline. Therefore, we can observe a gradual decrease in loss in the current output. In the next week, we can show more evaluate/train accuracy data based on our progress. We believe that the performance of our model meets the expected/planned standards.
Plan:
We need to spend more time on the final "model assembly/interface integration" to achieve a robust/integrated working pipeline. We plan to finish the pipeline development as soon as possible in order to begin experimenting with different sampling strategies.
Log in or sign up for Devpost to join the conversation.