Association of genomic subtypes of lower-grade gliomas with shape features automatically extracted by a deep learning algorithm

Abstract: Recent analysis identified distinct genomic subtypes of lower-grade glioma tumors which are associated with shape features. In this study, we propose a fully automatic way to quantify tumor imaging characteristics using deep learning-based segmentation and test whether these characteristics are predictive of tumor genomic subtypes. We used preoperative imaging and genomic data of 110 patients from 5 institutions with lower-grade gliomas from The Cancer Genome Atlas. Based on automatic deep learning segmentations, we extracted three features which quantify two-dimensional and three-dimensional characteristics of the tumors. Genomic data for the analyzed cohort of patients consisted of previously identified genomic clusters based on IDH mutation and 1p/19q co-deletion, DNA methylation, gene expression, DNA copy number, and microRNA expression. To analyze the relationship between the imaging features and genomic clusters, we conducted the Fisher exact test for 10 hypotheses for each pair of imaging feature and genomic subtype. To account for multiple hypothesis testing, we applied a Bonferroni correction. P-values lower than 0.005 were considered statistically significant. We found the strongest association between RNASeq clusters and the bounding ellipsoid volume ratio ($p<0.0002$) and between RNASeq clusters and margin fluctuation ($p<0.005$). In addition, we identified associations between bounding ellipsoid volume ratio and all tested molecular subtypes ($p<0.02$) as well as between angular standard deviation and RNASeq cluster ($p<0.02$). In terms of automatic tumor segmentation that was used to generate the quantitative image characteristics, our deep learning algorithm achieved a mean Dice coefficient of 82% which is comparable to human performance.

• The paper demonstrates that automated shape feature extraction correlates with genomic subtypes in lower-grade gliomas.
• The study employs a convolutional neural network to analyze imaging and genomic data, enhancing diagnostic precision.
• The findings indicate that integrating AI in radiology can refine personalized treatment strategies for glioma patients.

Analysis of Genomic Subtypes of Lower-Grade Gliomas Through Deep Learning-Extracted Shape Features

The paper explores the intersection of genomic data and radiological imaging in the context of lower-grade gliomas. It investigates the association between genomic subtypes of these brain tumors and shape features automatically extracted by a deep learning algorithm. This study is particularly relevant given the complexity and heterogeneity of gliomas, which pose significant challenges in diagnosis and treatment.

Methodology and Dataset

The research employs a dataset comprising imaging and genomic data associated with lower-grade gliomas. Utilizing a convolutional neural network (CNN), the study automates the extraction of shape features from imaging data. This approach illustrates the potential of deep learning technologies to handle the intricate relationships between medical imaging and genomic information, enabling more efficient feature extraction compared to traditional manual or semi-automated methods.

Results

The results of the study demonstrate a statistically significant association between automatically extracted shape features from imaging data and the genomic subtypes of lower-grade gliomas. The use of deep learning in this context highlights the capability of these models to discern subtle differences in image data that may correspond to underlying genomic profiles. Importantly, the extracted shape features achieved a noteworthy correlation with genomic subtypes, supporting the hypothesis that tumor morphology can reflect genomic variations.

Discussion and Implications

The implications of this research are multifaceted. Practically, the findings suggest that integrating automated imaging analysis with genomic data could enhance diagnostic accuracy and individualize treatment planning for glioma patients. Theoretically, this study contributes to the ongoing exploration of phenotypic-genotypic correlations, offering a computational approach to bridge the gap between radiology and genomics.

The paper also discusses the potential for future developments in AI within medical imaging and genomics. As deep learning algorithms continue to evolve, their ability to process and interpret complex datasets is expected to improve, potentially leading to more precise predictive models and a better understanding of cancer pathophysiology.

Limitations and Future Directions

Acknowledging the limitations of the study, the authors suggest that further research should include larger datasets and potentially incorporate additional imaging modalities to validate the generalizability of the findings. The robustness of the model could also benefit from incorporating diverse demographic and clinical data, which might unveil additional layers of complexity and enhance the semantic understanding of the extracted features.

In conclusion, this paper provides a significant step towards the application of deep learning models in extracting insightful features from medical images and associating them with genomic data. The research highlights a promising direction for the integration of AI in personalized medicine, offering a methodological framework that could be extended and refined in future studies for better clinical outcomes.