# Machine learning based classification of aggressive and malignant renal tumors from multimodal data

## TL;DR

This study uses machine learning with CT scans and clinical data to classify kidney tumors as benign, malignant-indolent, or malignant-aggressive, improving non-invasive diagnosis and treatment planning.

## Contribution

A novel machine learning pipeline that classifies renal tumors using multimodal data and identifies tumor aggressiveness for better clinical decision-making.

## Key findings

- The best models achieved an AUC of 0.90 for distinguishing aggressive from indolent tumors.
- Tumor size significantly improved classification accuracy when combined with CT and clinical data.
- Random forest and multi-layer perceptron models showed complementary strengths in different classification tasks.

## Abstract

This study aimed to develop and evaluate a machine learning pipeline using multiphase contrast-enhanced CT images and clinical data to classify renal tumors as benign, malignant-indolent, or malignant-aggressive, while assessing the contribution of each data source to the classification. In this retrospective study, 448 patients (mean age: 60.7 ± 12.6 years, 306 male, 142 female) who underwent nephrectomy and preoperative CECT between June 2008 and July 2018 were included. Tumors were histologically categorized as benign-indolent, malignant-indolent, or malignant-aggressive. Self-supervised feature extraction converted 4-phase CECT images into 512 real-valued features, combined with clinical data and tumor size for classification. Two machine learning classifiers, random forest (RF) and multi-layer perceptron (MLP), were used to predict tumor type. Nested five-fold cross-validation was employed for hyperparameter tuning and model evaluation, and performance was assessed using area under the curve (AUC) analysis. The best-performing models achieved an AUC of 0.90 (95% CI: 0.88–0.93) for classifying indolent versus aggressive tumors and 0.76 (95% CI: 0.71–0.81) for malignant versus benign tumors. Models incorporating tumor size significantly improved classification accuracy. RF classifiers excelled in distinguishing indolent from aggressive tumors, while MLP classifiers performed better for malignant versus benign classification. The machine learning pipeline demonstrated high accuracy in differentiating aggressive from indolent renal tumors, offering valuable prognostic insights for personalized treatment. Tumor size was a critical factor, complementing CECT images and clinical data. These findings highlight the potential of ML techniques in enhancing renal tumor risk stratification.

Non-invasive techniques for diagnosing cancerous kidney tumors are crucial since traditional diagnostic methods like biopsy and surgery can carry significant risks and burdens for the patient. By leveraging CT scans and routinely acquired clinical data, combined with the power of machine learning, we have developed a digital health capability that can accurately identify the type of kidney tumors without the need for invasive procedures. Traditionally, kidney tumors have been classified as malignant or benign. However, we recognize that classifying tumors as aggressive or indolent offers more practical value since it directly impacts the course of treatment prescribed for the patient. For example, if a tumor is classified as indolent, then the appropriate treatment might be much less aggressive and could involve an active surveillance regimen. To address this need, our approach classifies tumors based on both criteria; that is, malignant or benign, and aggressive or indolent. In the long run, by using this classifier, doctors may be able to make more informed decisions, prescribing the most effective treatments while minimizing risks and improving patient outcomes.