All researchProject overview

Medical imaging · 2023

Renal tumour classification and surgical approach modelling

Combining contrast-enhanced CT with clinical information to classify renal tumours and study how surgical approaches can be predicted.

MSc research at Qatar University

Pipeline diagram. CT slices are filtered to kidney slices, a YOLO model extracts the tumour region, and image features are combined with selected clinical features to classify tumour subtype and predict partial or radical nephrectomy.
Figure 1 from Mahmud et al. (2023). CT slices without kidney are removed before a detector extracts the tumour region. Image features and preoperative clinical features then feed the subtype and surgical-approach models.View full-size figureSource articleCC BY 4.0

Approach

Can CT features combined with clinical information classify renal tumours and predict the surgical approach?

The study used contrast-enhanced CT and metadata from the public KiTS cohort. It extracted a tumour region, learned imaging features, and combined those features with clinical variables. The study evaluated tumour subtype classification and prediction of partial versus radical nephrectomy as separate tasks.

  1. Extract the tumour region from CT
  2. Combine imaging and clinical features
  3. Classify tumours and predict surgery

Reported results

The paper reports 85.66% accuracy for the combined tumour classification model and 90.63% for surgical approach prediction in the evaluated cohort. Feature ranking identified tumour volume and cancer stage as informative for the surgical task.

Tumour subtype classification accuracy
85.66%
Surgical approach prediction accuracy
90.63%
Subjects in the KiTS21 cohort
300
Horizontal bar chart of random forest feature importance for surgical approach selection. Radiographic tumour size and cancer stage rank highest, followed by body mass index and age at nephrectomy.
Figure 9 from Mahmud et al. (2023). Random forest ranking of the clinical features that informed surgical approach prediction.View full-size figureSource articleCC BY 4.0

My contribution

I contributed to the study design, software, data curation, validation, visualisation and writing as part of my MSc thesis.

Study scope

The surgical model predicts approaches recorded in the study data. Prospective clinical validation would be needed before using it to guide care.