Cancer size, histotype, and cellular grade may limit the success of fine-needle aspiration cytology for screen-detected breast carcinoma

Evaluation of large language models in breast cancer clinical scenarios: a comparative analysis based on ChatGPT-3.5, ChatGPT-4.0, and Claude2

BACKGROUND

Large language models (LLMs) have garnered significant attention in the AI domain owing to their exemplary context recognition and response capabilities. However, the potential of LLMs in specific clinical scenarios, particularly in breast cancer diagnosis, treatment, and care, has not been fully explored. This study aimed to compare the performances of three major LLMs in the clinical context of breast cancer.

METHODS

In this study, clinical scenarios designed specifically for breast cancer were segmented into five pivotal domains (nine cases): assessment and diagnosis, treatment decision-making, postoperative care, psychosocial support, and prognosis and rehabilitation. The LLMs were used to generate feedback for various queries related to these domains. For each scenario, a panel of five breast cancer specialists, each with over a decade of experience, evaluated the feedback from LLMs. They assessed feedback concerning LLMs in terms of their quality, relevance, and applicability.

RESULTS

There was a moderate level of agreement among the raters (Fleiss' kappa=0.345, P<0.05). Comparing the performance of different models regarding response length, GPT-4.0 and GPT-3.5 provided relatively longer feedback than Claude2. Furthermore, across the nine case analyses, GPT-4.0 significantly outperformed the other two models in average quality, relevance, and applicability. Within the five clinical areas, GPT-4.0 markedly surpassed GPT-3.5 in the quality of the other four areas and scored higher than Claude2 in tasks related to psychosocial support and treatment decision-making.

CONCLUSION

This study revealed that in the realm of clinical applications for breast cancer, GPT-4.0 showcases not only superiority in terms of quality and relevance but also demonstrates exceptional capability in applicability, especially when compared to GPT-3.5. Relative to Claude2, GPT-4.0 holds advantages in specific domains. With the expanding use of LLMs in the clinical field, ongoing optimization and rigorous accuracy assessments are paramount.

Fluorescence Polarization of Methylene Blue as a Quantitative Marker of Breast Cancer at the Cellular Level

A quantitative technique to detect cancer in single cells could transform cancer diagnosis. Current cancer diagnosis utilizes histopathology, which requires tissue acquisition, extensive processing and, in most cases, relies on the qualitative morphological analysis of tissues and cells. Molecular biomarkers are only available for a few specific tumor subtypes. We discovered that the fluorescence polarization (Fpol) of Methylene Blue (MB) is significantly higher in cancer than in normal human breast tissues and cells. We confirmed that fluorescence polarization imaging did not affect the viability of the cells and yielded highly significant differences between cancer and normal cells using MB concentrations as low as 0.05 and 0.01 mg/ml. To explain this phenomenon we examined intracellular localization of MB and its fluorescence lifetime. We determined that higher fluorescence polarization of MB occurs due to its increased accumulation in mitochondria of cancer cells, as well as shorter fluorescence lifetime in cancer relative to normal cells. As quantitative MB Fpol imaging can be performed in vivo and in real time, it holds the potential to provide an accurate quantitative marker of cancer at the cellular level.

Fine needle aspiration cytology of lobular breast carcinoma and its variants

OBJECTIVE

To identify associations between cytological criteria in fine needle aspiration (FNA) specimens and histological subtypes of lobular breast carcinoma (classical and other types).

STUDY DESIGN

FNA cytology and mastectomy specimens from 72 cases of invasive lobular breast carcinoma were consecutively retrieved from the files of the Amaral de Carvalho Hospital, Jaú-São Paulo, Brazil. All cases were reviewed regarding five cytological criteria: cellularity, cellular cohesion, presence of inflammation, nucleoli and nuclear atypia. The χ2 test or Fisher's exact tests with 95% confidence intervals (CI) were used.

RESULTS

The classical type showed lower initial cytological diagnosis of malignancy compared to the other variants (p=0.017; odds ratio (OR) 0.26, 95% CI 0.89-0.80). Moderate/intense cellular cohesion (p=0.011; OR 0.18, 95% CI 0.04-0.73) and mild atypia (p=0.000; OR 16.15, 95% CI 3.20-81.48) were significantly associated with the classical type of lobular breast carcinoma, while the absence of inflammation (p=0.082; OR 0.36, 95% CI 0.12-1.15) was marginally associated with the classical type.

CONCLUSIONS

In cytology, the characterization of lobular carcinoma as malignant is difficult, especially the classical type. The association between cell cohesion and the classical type of lobular breast carcinoma may be one of the factors that complicate this diagnosis.

The microscopic complexities of C3 in breast cytology

BACKGROUND

Fine-needle aspiration (FNA) of difficult breast lesions often results in an atypical (C3) report. The assortment of outcomes generated by C3 reports varies widely, and this has given rise to different clinical management pathways.

OBJECTIVE

To identify and objectively assess microscopic features associated with atypical/C3 breast FNA cases.

MATERIALS AND METHODS

A total of 230 atypical breast FNAs were subjected to a blind microscopic rescreen using a range of robust qualitative and quantitative cytological criteria including cellularity, architectural qualities, cytomorphology and background features. A logistic regression with a receiver operating characteristic (ROC) curve and the resultant forward stepwise analysis were conducted to assess the results. This statistical testing was measured against malignant, benign proliferative and benign non-proliferative outcomes.

RESULTS

The malignant and benign proliferative outcomes showed a mixture of opposing protective and predictive individual cytological criteria. The stepwise analysis produced models demonstrating the best combination of individual cytological criteria for malignancy, proliferative and benign non-proliferative entities. In the malignancy model, discohesion, nuclear crowding within sheets, diminished numbers of bare bipolar nuclei and myoepithelial cells, the presence of tubules or necrosis and the absence of a cystic background were important features. The benign proliferative model suggested the same criteria but with the opposite implication and with the addition of several others, such as the presence of apocrine metaplasia, retained polarity and a speckled or coarse chromatin pattern. Age was a significant factor in malignant and proliferative outcomes. The benign non-proliferative stepwise analysis produced a model with fewer criteria (complex sheets, bare bipolar nuclei and a cystic background) limiting clinical application.

CONCLUSION

Atypical/C3 breast cytology remains a legitimate reporting category. However, it is associated with a number of different histological outcomes. The incidence of the C3 category can be significantly reduced by controlling extrinsic factors and understanding the associated microscopic features.

The true nature of atypical breast cytology

BACKGROUND

Atypical breast cytology is a poorly understood heterogeneous category with limited clinical utility but significant implications for patient management.

OBJECTIVE

To provide an insight into the true nature of atypical breast cytology in screening-detected (asymptomatic) and symptomatic settings, and find strategies for reducing the use of this diagnostic category.

MATERIALS AND METHODS

A total of 6,415 breast cytology samples were processed between January 2004 and December 2008. An atypical cytological diagnosis was rendered in 256 (4%) of the cases. A blind microscopic review of the atypical cases was conducted and results were correlated with subsequent histological and/or clinical outcomes.

RESULTS

Follow-up information by histology was available in 85.5%, by repeat fine-needle aspiration (FNA) in 3.5% and by imaging or clinical follow-up in 10.2% of the cases. Two patients (0.8%) were lost to follow-up. Of the 254 cases with follow-up, 62.6% were benign and 37.4% were malignant. The benign to malignant ratios were 1:1 and 2:1 in the screening and symptomatic groups, respectively. The atypical category in the screening population mostly yielded fat necrosis, complex sclerosing lesions and low- to intermediate-grade carcinoma on follow-up. The main outcomes in the symptomatic group were papilloma, fibroadenoma, ductal carcinoma in situ and lobular carcinoma. Preanalytical (suboptimal samples) factors were encountered in 34.8% and interpretative factors in 65.2% of the cases. Uncertainty about cellular morphology was attributed to such a diagnosis in 38 (14.8%) of the cases, architectural complexity in 137 (53.5%) and morphology and architecture in 70 (27.3%); 4.3% of cases were considered nondiagnostic.

CONCLUSION

The atypical category is a necessary diagnosis but of limited use from a patient management perspective. Some preanalytical factors such as poor sample quality can be minimized by the involvement of cytopathologists in the FNA procedure. The use of the atypical category is partly dependent on the experience and confidence of the reporting pathologist. Assigning a case to this category is also likely to be unduly influenced by clinical or radiological findings. Our study indicates that the use of the atypical category can be reduced by up to 40% by appreciating these contributing factors. The practical utilization of the atypical category in breast cytology remains subjective and further study is required to identify useful objective criteria.

Ultrasound-guided vacuum-assisted core biopsy in the diagnosis and treatment of focal lesions of the breast - own experience

Introduction

Vacuum-assisted core biopsy (VACB) guided by ultrasound is a minimally invasive method used in diagnosis and treatment of breast focal lesions. Vacuum-assisted core biopsy is an interesting minimally invasive alternative to open surgical biopsy.

Aim

To assess the value of ultrasound-guided vacuum-assisted core biopsy in the diagnosis and treatment of breast focal lesions.

Material and methods

In the period 2009-2010, 397 ultrasound-guided vacuum-assisted core biopsies were performed. Mean age of the patients was 41.7 years (18-92 years), and size of the lesions ranged from 3 mm to 65 mm, mean size being 12 mm. All women with diagnosed atypical ductal hyperplasia or cancer were qualified for surgery. The patients with histopathologically benign lesions were under follow-up.

Results

Samples sufficient for histopathological examination were obtained from 394 cases (99.2%). Of all 397 lesions, 293 (73.7%) were diagnosed as benign, there were 6 (1.6%) cases of atypical ductal hyperplasia and 98 (24.7%) malignant lesions. Three hundred and sixty-nine lesions were below 15 mm in diameter, of which 339 (91.9%) were totally removed during the VACB.

Conclusions

The results obtained confirm high efficiency of ultrasound-guided VACB in the differential diagnosis of breast focal lesions, including impalpable ones. It is a safe method with a low complication rate. In the case of benign lesions with a diameter not exceeding 15 mm, it allows one to excise the whole lesion and is a very good alternative to an open surgical biopsy. Vacuum-assisted core biopsy should be a standard and the method of choice in diagnosing breast lesions.