Toward New Assessment in Sarcoma Identification and Grading Using Artificial Intelligence Techniques

Abstract

Background/Objectives: Sarcomas are a rare and heterogeneous group of malignant tumors, which makes early detection and grading particularly challenging. Diagnosis traditionally relies on expert visual interpretation of histopathological biopsies and radiological imaging, processes that can be time-consuming, subjective and susceptible to inter-observer variability. Methods: In this study, we aim to explore the potential of artificial intelligence (AI), specifically radiomics and machine learning (ML), to support sarcoma diagnosis and grading based on MRI scans. We extracted quantitative features from both raw and wavelet-transformed images, including first-order statistics and texture descriptors such as the gray-level co-occurrence matrix (GLCM), gray-level size-zone matrix (GLSZM), gray-level run-length matrix (GLRLM), and neighboring gray tone difference matrix (NGTDM). These features were used to train ML models for two tasks: binary classification of healthy vs. pathological tissue and prognostic grading of sarcomas based on the French FNCLCC system. Results: The binary classification achieved an accuracy of 76.02% using a combination of features from both raw and transformed images. FNCLCC grade classification reached an accuracy of 57.6% under the same conditions. Specifically, wavelet transforms of raw images boosted classification accuracy, hinting at the large potential that image transforms can add to these tasks. Conclusions: Our findings highlight the value of combining multiple radiomic features and demonstrate that wavelet transforms significantly enhance classification performance. By outlining the potential of AI-based approaches in sarcoma diagnostics, this work seeks to promote the development of decision support systems that could assist clinicians.

Keywords: classification; image transform; machine learning; radiomics; sarcoma.

Figure 1

Visual representation of the steps in the workflow.

Figure 2

Confusion matrix setup. (a) For binary classification, true negative (TN) indicates tissue correctly predicted as healthy, true positive (TP) indicates tissue correctly predicted as pathological, false negative (FN) indicates pathological tissue predicted as healthy and false positive (FP) indicates pathological tissue predicted as healthy. The micro average displays the number of samples classified in integers, and macro average displays classification in ratios. (b) For grade classification, true grade ($T{G}_i$) is the number of samples belonging to class i and predicted as such, and false grade ($F{G}_{i,j}$) corresponds to samples predicted as grade i but belonging to grade j. The micro average shows the total number of classified data points, and the macro average shows number of classified patients.

Figure 3

MRI slices: (a) A case of OS where the machine uses specific settings and contrast media to brighten the tumor region and suppress fat response. (b) Another patient (with liposarcoma in this case) where the tumor region is dark in contrast and fat tissue is not suppressed.

Figure 4

An example of how healthy and pathological tissue can overlap in terms of appearance. The left two images show healthy tissue, and the rightmost image shows a tumor region.

Figure 5

Variance in image quality. (a) A grainy slice with relatively high noise. (b) A corrupted image. (c) An example of uneven lighting distribution.