Quantitative evaluation of Saliency-Based Explainable artificial intelligence (XAI) methods in Deep Learning-Based mammogram analysis.

BACKGROUND: Explainable Artificial Intelligence (XAI) is prominent in the diagnostics of opaque deep learning (DL) models, especially in medical imaging. Saliency methods are commonly used, yet there's a lack of quantitative evidence regarding their performance. OBJECTIVES: To quantitatively evaluate the performance of widely utilized saliency XAI methods in the task of breast cancer detection on mammograms. METHODS: Three radiologists drew ground-truth boxes on a balanced mammogram dataset of women (n = 1496 cancer-positive and negative scans) from three centers. A modified, pre-trained DL model was employed for breast cancer detection, using MLO and CC images. Saliency XAI methods, including Gradient-weighted Class Activation Mapping (Grad-CAM), Grad-CAM++, and Eigen-CAM, were evaluated. We utilized the Pointing Game to assess these methods, determining if the maximum value of a saliency map aligned with the bounding boxes, representing the ratio of correctly identified lesions among all cancer patients, with a value ranging from 0 to 1. RESULTS: The development sample included 2,244 women (75%), with the remaining 748 women (25%) in the testing set for unbiased XAI evaluation. The model's recall, precision, accuracy, and F1-Score in identifying cancer in the testing set were 69%, 88%, 80%, and 0.77, respectively. The Pointing Game Scores for Grad-CAM, Grad-CAM++, and Eigen-CAM were 0.41, 0.30, and 0.35 in women with cancer and marginally increased to 0.41, 0.31, and 0.36 when considering only true-positive samples. CONCLUSIONS: While saliency-based methods provide some degree of explainability, they frequently fall short in delineating how DL models arrive at decisions in a considerable number of instances.

Retrospective Analysis of Parameters Affecting Metastatic Breast Cancer.

Objectives: While metastatic breast cancer (MBC), which is the most common cause of death in women, has been seen as an incurable surgical problem in the past decade, as the heterogeneous nature of breast cancer becomes clear with increasing molecular studies and advances in oncological protocols, life expectancy is increasing. In this study, we aimed to examine the clinicopathological features of the patients we followed up with MBC. Methods: Patients who were operated on with the diagnosis of breast cancer in our hospital between 2018 and 2023 and who were later found to have metastases were retrospectively analyzed from the database. The age of the patients, the histological and molecular type, stage and grade of the tumor, the time from diagnosis to metastasis, the location of metastasis, the duration of treatment and follow-up were investigated. Patients who were operated on in other centers and/or were out of follow-up were excluded from the study. For the statistical analysis of the findings, number cruncher statistical system (NCSS) 2020 statistical software (NCSS LLC, Kaysville, Utah, USA) was used at a significance level of 0.05. Results: Metastasis was detected in 77.1% (n=37) of a total of 48 female patients, and recurrence was found in 22.9% (n=11). The mean age of the patients was 57 years. There was no statistically significant difference between the patients in terms of demographics. When evaluated according to the TNM stage, 24.3% (n=9) of the patients were in the early stage and 75.7% (n=28) were in the locally advanced stage; the number of locally advanced patients was found to be higher than the early stage. In histology examination, 27.1% (n=13) of the patients were luminal A, 31.3% (n=15) luminal B, 16.7% (n=8) HER2 positive, and 25% (n=12) triple negative. Ki67 was higher than 14% in 64.6% (n=31) patients. Breast conserving surgery was performed in 41.6% (n=20) of the patients, and mastectomy was performed in 58.3% (n=28) patients. Metastasis in 34.2% (n=13) of the cases within 1-2 years, in 42.1% (n=16) within 2-5 years, and in 23.7% (n=9) after 5 years took place. Sites of metastasis were bone (37.7%, n=28), liver (28.9%, n=11), brain (10.5%, n=4), and lung (7.9%, n=3). More than one metastasis site was observed in 21.05% (n=8) of patients with metastases. There was no statistically significant difference between luminal A, luminal B, HER 2 groups and triple-negative breast cancer in terms of metastasis time and location (p>0.05). Adjuvant hormone therapy was more common in the luminal A group, whereas neoadjuvant therapy was more common in the HER2+ group. A total of 20 deaths were observed in 48 patients (41.7%). The median disease-free survival was 64 months. Conclusion: Despite all the developments in metastatic breast cancer, the 5-year survival rate is 27%. Targeted personalized therapies may be promising when the mechanism of metastasis and specific pathways in breast cancer emerge.

Feasibility and accuracy of targeted axillary dissection in breast cancer patients; single center experience.

Introduction: Axillary complete response (pCR) was observed in approximately half of breast cancer patients who received neoadjuvant chemotherapy (NAC) due to axillary positivity. Preventing axillary morbidity due to unnecessary axillary lymph node dissection (ALND) is extremely important for patients' quality of life. Targeted axillary dissection (TAD) is a technique developed to improve axillary staging and reduce the false negative rate in sentinel lymph node biopsy. Methods: Patients with cN1-N2 (clinically node) breast cancer whose axillary lymph node metastasis was confirmed by biopsy and who received NAC marked with a clip were included in the study. Patients who achieved clinical response after systemic treatment underwent sentinel lymph node biopsy (SLNB) with additional methods such as methylene blue guided dissection, skin marking for clip on lymph node (CLN) localization, and wire guided with imaging excision of the CLN. TAD and ALND pathology results were evaluated and analyzed with patient and tumor characteristics. Results: 83 patients who met the eligibility criterias for the study were examined retrospectively. 21 of the patients underwent TAD alone, and 62 patients underwent ALND after TAD. CLN rate was 98.8% in patients underwent only TAD and this rate was increased to 100% in patients underwent ALND after TAD. FNR in SLN, CLN, and TAD were 28.6%, 10.7%, and 3.6%, respectively. Conclusion: TAD is a feasible and reliable surgical approach to detect axillary residual disease with a high success rate.

Simple Blood Tests May Be Used to Predict the Increased Risk of Conversion in Elective Laparoscopic Cholecystectomy Surgery.

Background: In this study, we aimed to evaluate the possibility if simple blood tests that can be made in majority of hospitals may be used predict to risk of conversion to laparoscopic surgery to an open approach. Patients and Methods: The hospital records of 636 patients who underwent elective laparoscopic cholecystectomy (L-C) were retrospectively reviewed, and 583 patients included in the study protocol. Preoperative laboratory tests of all patients and data of patients who underwent conversion from laparoscopic surgery to open surgery were examined. Results: Of the 583 patients who were included in the study, 404 (69.29%) were female and the mean age was 50.02 ± 12.84 (19-89) years. The cholecystectomy was completed laparoscopically in 559 (89.5%) patients. The most common symptoms seen in the patients were epigastric discomfort and right upper quadrant pain. The high level of white blood cell (WBC) count and c-reactive protein (CRP) were found to be statistically significant before surgery in patients who had a conversion to open cholecystectomy (P < .001). Conclusion: Elevation of WBC count and CRP value before elective L-C may be useful in the prediction of a high risk of conversion from laparoscopic to open approach. This finding will help the surgeon to plan the treatment and inform the patient of the possibility before surgery.

Efficacy of Preoperative Imaging Features and Blood Tests in Predicting the Increased Risk of Conversion in Laparoscopic Appendectomy Surgery.

Background In this study, we identified preoperative risk factors, including imaging features and blood tests, to predict conversion from laparoscopic appendectomy to open appendectomy. Thus, we aimed to prevent patients from being exposed to the risks of laparoscopy by choosing patients for whom proceeding directly to an open surgery as an initial approach was appropriate. Patients and methods The cohort of 632 patients who underwent laparoscopic appendectomy due to acute appendicitis in our center between January 2017 and March 2021 were analyzed, and 521 of these patients comprised the study population. Baseline characteristics, medical history, preoperative laboratory tests, imaging features, and postoperative pathologic findings of all patients according to groups who underwent laparoscopic appendectomy or conversion to open appendectomy were examined. Results Among 521 patients, the appendectomy procedure was completed laparoscopically in 498 (95.6%) patients, and conversion to open appendectomy was occurred in 23 (4.4%) patients. 223 (42.8%) patients were female, and 298 (57.2%) patients were male. The mean age of all patients was 35.17±12.61 years (range, 16-80 years). Preoperative ultrasonography feature associated with a higher rate of conversion was free fluid collection (p=0.001). The levels of C-reactive protein, neutrophil, and neutrophil/lymphocyte ratio on admission were found to be significantly higher in the conversion group compared to the laparoscopy group (p=0.001, p=0.027, p=0.02, respectively).  Conclusions Free fluid collection detected by ultrasonography, and elevation of C-reactive protein, neutrophil, and neutrophil/lymphocyte ratio may be useful in the prediction of a high risk of conversion appendectomy. Despite the unquestionable advantages of laparoscopic surgery, there are still substantial conversion rates. Within this framework, our study will help the surgeons to choose the most appropriate surgical methods for patients by evaluating them individually, and to inform them of the possibility of conversion to the open approach, and other risks before surgery.