JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

Whole-body low-dose computed tomography in patients with newly diagnosed multiple myeloma predicts cytogenetic risk: a deep learning radiogenomics study

  • 0Department of Radiology, Mayo Clinic, 200 1st St SW, Rochester, MN, 55905, USA.
Skeletal Radiology +

|

June 27, 2024

|

June 27, 2024

View abstract on PubMed

Summary

This summary is machine-generated.

A new deep learning model using whole-body low-dose CT scans accurately predicts cytogenetic abnormalities in multiple myeloma (MM). This tool aids in risk stratification and survival prediction for MM patients.

Area Of Science

  • Radiology and Medical Imaging
  • Oncology
  • Artificial Intelligence in Medicine

Background

  • Cytogenetic abnormalities are crucial for risk stratification in multiple myeloma (MM).
  • Accurate prediction of these abnormalities can improve patient management and survival outcomes.
  • Whole-body low-dose CT (WBLDCT) scans offer a comprehensive imaging approach.

Purpose Of The Study

  • To develop a deep learning (DL) model utilizing WBLDCT scans.
  • To assess the DL model's accuracy in predicting cytogenetic abnormalities in MM.
  • To evaluate the model's utility in risk stratification and survival prediction.

Main Methods

  • WBLDCT scans from 151 MM patients within one year of diagnosis were analyzed.
  • Fluorescent in situ hybridization (FISH) was used for cytogenetic assessment and risk stratification (high-risk vs. standard-risk).
  • A DL model was trained and validated using a five-fold cross-validation approach, with Area Under the Receiver Operating Curve (AUROC) as the performance metric.

Main Results

  • The DL model demonstrated good to excellent performance in classifying various cytogenetic abnormalities.
  • The highest AUROC was observed for t(4;14) (0.874 ± 0.073), while trisomies had the lowest AUROC (0.717 ± 0.058).
  • The model's survival predictions aligned closely with actual survival rates for both high-risk and standard-risk groups.

Conclusions

  • A WBLDCT-based DL model can effectively predict cytogenetic abnormalities in MM.
  • This model shows promise for improving risk stratification and guiding treatment decisions in MM patients.
  • The findings highlight the potential of AI in medical imaging for oncological applications.

Keywords: