Predicting neo‐adjuvant chemotherapy response and progression‐free survival of locally advanced breast cancer using textural features of intratumoral heterogeneity on F‐18 FDG PET / CT and diffusion‐weighted MR imaging

Abstract

Predicting response to neo-adjuvant chemotherapy (NAC) and survival in locally advanced breast cancer (LABC) is important. This study investigated the prognostic value of tumor heterogeneity evaluated with textural analysis through F-18 fluorodeoxyglucose (FDG) positron emission tomography (PET) and diffusion-weighted imaging (DWI). We enrolled 83 patients with LABC who had completed NAC and curative surgery. Tumor texture indices from pretreatment FDG PET and DWI were extracted from histogram analysis and 7 different parent matrices: co-occurrence matrix, the voxel-alignment matrix, neighborhood intensity difference matrix, intensity size-zone matrix (ISZM), normalized gray-level co-occurrence matrix (NGLCM), neighboring gray-level dependence matrix (NGLDM), and texture spectrum matrix. The predictive values of textural features were tested regarding both pathologic NAC response and progression-free survival. Among 83 patients, 46 were pathologic responders, while 37 were nonresponders. The PET texture indices from 7 parent matrices, DWI texture indices from histogram, and 1 parent matrix (NGLCM) showed significant differences according to NAC response. On multivariable analysis, number nonuniformity of PET extracted from the NGLDM was an independent predictor of pathologic response (P = .009). During a median follow-up period of 17.3 months, 14 patients experienced recurrence. High-intensity zone emphasis (HIZE) and high-intensity short-zone emphasis (HISZE) from PET extracted from ISZM were significant textural predictors (P = .011 and P = .033). On Cox regression analysis, only HIZE was a significant predictor of recurrence (P = .027), while HISZE showed borderline significance (P = .107). Tumor texture indices are useful for NAC response prediction in LABC. Moreover, PET texture indices can help to predict disease recurrence.

Machine Learning-Based Radiomics Nomogram Using Magnetic Resonance Images for Prediction of Neoadjuvant Chemotherapy Efficacy in Breast Cancer Patients

“…However, in this study, we created a pairwise sorting combination of three sequences (T2WI, DWI, and DCE), which further expanded the scope of this type of research. In addition, our result is consistent with Yoon's results that the features most associated with the clinical outcome of pCR mainly include Inverse difference moment and short-run emphasis features (28), which indicated these two types of feature had significant difference between non-pCR patients and pCR patients with breast cancer experiencing NACT. Although the remaining features also represent the degree of heterogeneity, the exact calculation method of each feature value varies depending on the parameters.…”

The clinical utility of prostate cancer heterogeneity using texture analysis of multiparametric MRI

“…Increased tumour heterogeneity has been previously linked to worse clinical prognosis in multiple tumour types, including oesophageal, colorectal, CNS, and non-small cell lung carcinoma. Furthermore, texture analysis can provide prognostic information from imaging, additional to that given by a radiologist, and be used as an independent predictor of survival [19,30]. Texture analysis has also been harnessed in breast, rectal, renal cell and pancreatic cancer as an early indicator of treatment response [10][11][12][13][14][15]30].…”

MRI-Based Radiomics Analysis for the Pretreatment Prediction of Pathologic Complete Tumor Response to Neoadjuvant Systemic Therapy in Breast Cancer Patients: A Multicenter Study

“…Several MRI-based radiomics studies have reported promising results regarding the prediction of pCR to NST in breast cancer patients based on pretreatment scans [ 17 , 18 , 19 , 20 , 21 ]. However, the evidence from these studies is limited due to the relatively small sample sizes ranging from 29 to 100 patients and the lack of external validation datasets.…”