According to the American Cancer Society, more than 229,000 people will be diagnosed with lung cancer in the United States this year, with adenocarcinoma being the most common type. To help with diagnosis, researchers from Dartmouth’s Norris Cotton Cancer Center and the Hassanpour Lab at Dartmouth University developed a deep learning-based system for automated classification of histologic subtypes on lung adenocarcinoma surgical resection slides on par with pathologists.
“Classification of histologic patterns in lung adenocarcinoma is critical for determining tumor grade and treatment for patients,” the researchers stated in their paper. “Our model can potentially be used to aid pathologists in the classification of these histologic patterns and ultimately contribute to more accurate grading of lung adenocarcinoma,” the team said.
Using NVIDIA Tesla GPUs with the cuDNN-accelerated PyTorch deep learning framework, the team trained a convolutional neural network on hundreds of whole-slide images from patients with a diagnosis of lung adenocarcinoma who underwent lobectomies at the Dartmouth-Hitchcock Medical Center (DHMC).
“Our model uses a convolutional neural network to identify regions of neoplastic cells, then aggregates those classifications to infer predominant and minor histologic patterns for any given whole-slide image,” the researchers said.
Once the neural network was trained, the team evaluated the algorithm on a dataset of 143 images. According to the team, all evaluation metrics for the model were within 95% confidence intervals of agreement with the pathologist’s assessment.
“Considering the quick turnaround time of our model, it could be integrated into existing laboratory information management systems to automatically pre-populate diagnoses for histologic patterns on slides or provide a second opinion on more challenging patterns. In addition, a visualization of the entire slide, examined by our model at the piecewise level, could highlight elusive areas of high-grade patterns as well as primary regions of tumor cells,” the researchers stated. The application of our model in a clinical setting, which our research team will pursue as a next step, is essentially an automated platform for quality assurance in reading histologic slides of lung adenocarcinoma. A successful implementation of this system will support more accurate classification of lung cancer grade and ultimately facilitate the entire process of lung cancer diagnosis,” the team said.
The work was recently published in Nature, Scientific Reports in a paper titled Pathologist-level Classification of Histologic Patterns on Resected Lung Adenocarcinoma Slides with Deep Neural Networks.
The researchers also published a practical how-to blog explaining how you can train your own deep learning-based classifier on histopathology images. The code used in the paper is also available on GitHub.
