Scholarlink

As the days are moving forward, we are getting more engaged with AI. Nowadays, we are applying AI into everything, even in the medical sector. AI can be implemented in the medical sector in such a way: we can identify a disease with a trained model, we can also suggest treatments or medicines to the patient. We worked with the former. Our goal is to train a model in such a way that, after receiving histopathology images, it would be able to find cancer cells from that image, making the process of identifying the cancer cells more accurate and easier. We used VGG19, DenseNet201 and ResNet50 for our work. We experimented with pre-trained models, custom models and we also used the k-fold method to find out which model performs best. We used only lung images from the LC25000 dataset for this purpose, where the images were preprocessed. After our experimentation, we found that ResNet50 as a pretrained model performs best compared with the other two models, VGG19 as our custom model performs better than the other two and DenseNet201 performs well in terms of the k-fold method.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Histopathology Images-Based Deep Learning Prediction of Prognosis and Therapeutic Response in Small Cell Lung Cancer

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Karib Shams,
  • Aritra Das,
  • Md. Rakibul Hasan,
  • Mithila Sultana,
  • Md. Ahnaf Morshed,
  • Ahmed Wasif Reza

摘要

As the days are moving forward, we are getting more engaged with AI. Nowadays, we are applying AI into everything, even in the medical sector. AI can be implemented in the medical sector in such a way: we can identify a disease with a trained model, we can also suggest treatments or medicines to the patient. We worked with the former. Our goal is to train a model in such a way that, after receiving histopathology images, it would be able to find cancer cells from that image, making the process of identifying the cancer cells more accurate and easier. We used VGG19, DenseNet201 and ResNet50 for our work. We experimented with pre-trained models, custom models and we also used the k-fold method to find out which model performs best. We used only lung images from the LC25000 dataset for this purpose, where the images were preprocessed. After our experimentation, we found that ResNet50 as a pretrained model performs best compared with the other two models, VGG19 as our custom model performs better than the other two and DenseNet201 performs well in terms of the k-fold method.