Validation of artificial intelligence contrast mammography in diagnosis of breast cancer: Relationship to histopathological results

INTRODUCTION

Contrast-enhanced mammography (CEM) is used for characterization of breast lesions with increased diagnostic accuracy compared to digital mammography (DM). Artificial intelligence (AI) approaches are emerging with accuracies equal to an average radiologist. However, most studies trained deep learning (DL) models on DM images and there is a paucity in literature for discovering the application of AI using CEM.

OBJECTIVES

To develop and test a DL model that classifies CEM images and produces corresponding highlights of lesions detected.

METHODS

Fully annotated 2006 images of 326 females available from the previously published Categorized Digital Database for Contrast Enhanced Mammography images (CDD-CESM) were used for training. We developed a DL multiview contrast mammography model (MVCM) for classification of CEM low energy and recombined images. An external test set of 288 images of 37 females not included in the training was used for validation. Correlation with histopathological results and follow-up was considered the standard reference. The study protocol was approved by the Institutional Review Board and patient informed consent was obtained.

RESULTS

Assessment was done on an external test set of 37 females (mean age, 51.31 years ± 11.07 [SD]) with AUC-ROC for AI performance 0.936; (95 % CI: 0.898, 0.973; p < 0.001) and the best cut off value for prediction of malignancy using AI score = 0.28. Findings were then correlated with histopathological results and follow up which revealed a sensitivity of 75 %, specificity 96.3 %, total accuracy of 90.1 %, positive predictive value (PPV) 87.1 %, and negative predictive value (NPV) 92 %, p-value (<0.001). Diagnostic indices of radiologists were sensitivity 88.9 %, specificity 92.6 %, total accuracy 91.7 %, PPV 80 %, and NPV 96.2 %, p-value (<0.001).

CONCLUSION

A deep learning multiview CEM model was developed and evaluated in a cohort of female participants and showed promising results in detecting breast cancer. This warrants further studies, external training, and validation.

Does Gelfoam slurry embolization post-pulmonary biopsy reduce risk of pneumothorax? A prospective randomized control study

Background

CT-guided percutaneous transthoracic biopsy has become a widely accepted safe method in establishing the etiology of lung masses. Pneumothorax is the most commonly faced complication requiring further therapeutic intervention for treatment with chest tube drainage adding to the time of hospital stay.

Aim of work

We aim to evaluate the effect of Gelfoam mixture biopsy tract embolization as a minimal added cost in reducing the overall rate of complications, the need of chest tube placement, and hospital stay.

Patient and methods

A total of 138 transthoracic CT-guided lung biopsies were randomized to 70 track embolized patients and 68 control. The study protocol was approved from the National Cancer Institute ethical committee review board. Data were collected from the local PACS system and analyzed, reviewed and performed by one of three interventional radiologists. Patient records were analyzed for primary health risks, pre-procedural CT was analyzed for lesion-related risk factors, and intra-procedural CTs were analyzed for procedure-related risk factors. Outcome measures include incidence of pneumothorax, pneumothoraxes requiring chest tube insertion and hospital admission rates.

Results

Marked pneumothorax requiring chest tube insertion and hospital admission were significantly reduced in the embolized group as 7.1% compared to 19% in the non-embolized group with significant p value of 0.037. Hospital admissions reduced from 19% down to 4.3% in embolized cases with a significant p value of 0.007. On univariate regression analysis, embolization reduced chest tube placement odds by 68% (OR = 0.32, 95% CI 0.109–0.97, p = 0.044). The only significant procedural-related factor was needle pleural angle > 70° where it increased the risk of pneumothorax by 2.85 times and chest tube placement by 3.10 times. Gelfoam embolization significantly reduces the odds of post-procedural hospital admission by 81% (OR = 0.189, 95% CI 0.051–0.699, p = 0.012). In multivariate regression analysis, Gelfoam was significantly protective against chest tube insertion and prolonged hospital stay, by reducing the odds 74.3% (OR = 0.257, 95% CI 0.082–0.808, p = 0.020) and 86% (OR = 0.133, 95% CI 0.027–0.662, p = 0.014), respectively. Needle pleural angle more than 70° increased odds of chest tube insertion by 252%. Lesions that were in very low position related to the diaphragm and just behind ribs were less prone to chest tube insertion by 83% (OR = 0.164, 95% CI 0.035–0.779, p = 0.02), while those showing mediastinal invasion and central lung lesions had increased odds by 6.812 times (95% CI 1.452–31.958, p = 0.015) for longer hospital stays.

Conclusions

Gelfoam embolization post-CT-guided lung biopsy has proven to statistically reduce large pneumothoraces requiring chest tube insertion, longer hospital stay, and reduced unnecessary expenses.

Categorized contrast enhanced mammography dataset for diagnostic and artificial intelligence research

Contrast-enhanced spectral mammography (CESM) is a relatively recent imaging modality with increased diagnostic accuracy compared to digital mammography (DM). New deep learning (DL) models were developed that have accuracies equal to that of an average radiologist. However, most studies trained the DL models on DM images as no datasets exist for CESM images. We aim to resolve this limitation by releasing a Categorized Digital Database for Low energy and Subtracted Contrast Enhanced Spectral Mammography images (CDD-CESM) to evaluate decision support systems. The dataset includes 2006 images, with an average resolution of 2355 × 1315, consisting of 310 mass images, 48 architectural distortion images, 222 asymmetry images, 238 calcifications images, 334 mass enhancement images, 184 non-mass enhancement images, 159 postoperative images, 8 post neoadjuvant chemotherapy images, and 751 normal images, with 248 images having more than one finding. This is the first dataset to incorporate data selection, segmentation annotation, medical reports, and pathological diagnosis for all cases. Moreover, we propose and evaluate a DL-based technique to automatically segment abnormal findings in images.

Measurement(s)	Dual-Energy Contrast-Enhanced Digital Spectral Mammography
Technology Type(s)	digital curation
Sample Characteristic - Organism	Homo sapiens • Breast
Sample Characteristic - Location	Egypt