A Matched-Pair Analysis of Nuclear Morphologic Features Between Core Needle Biopsy and Surgical Specimen in Thyroid Tumors Using a Deep Learning Model

Artificial intelligence model predicts M2 macrophage levels and HCC prognosis with only globally labeled pathological images

Background and aims

The levels of M2 macrophages are significantly associated with the prognosis of hepatocellular carcinoma (HCC), however, current detection methods in clinical settings remain challenging. Our study aims to develop a weakly supervised artificial intelligence model using globally labeled histological images, to predict M2 macrophage levels and forecast the prognosis of HCC patients by integrating clinical features.

Methods

CIBERSORTx was used to calculate M2 macrophage abundance. We developed a slide-level, weakly-supervised clustering method for Whole Slide Images (WSIs) by integrating Masked Autoencoders (MAE) with ResNet-32t to predict M2 macrophage abundance.

Results

We developed an MAE-ResNet model to predict M2 macrophage levels using WSIs. In the testing dataset, the area under the curve (AUC) (95% CI) was 0.73 (0.59-0.87). We constructed a Cox regression model showing that the predicted probabilities of M2 macrophage abundance were negatively associated with the prognosis of HCC (HR=1.89, p=0.031). Furthermore, we incorporated clinical data, screened variables using Lasso regression, and built the comprehensive prediction model that better predicted prognosis. (HR=2.359, p=0.001).

Conclusion

Our models effectively predicted M2 macrophage levels and HCC prognosis. The findings suggest that our models offer a novel method for determining biomarker levels and forecasting prognosis, eliminating additional clinical tests, thereby delivering substantial clinical benefits.

A novel computational pathology approach for identifying gene signatures prognostic of disease-free survival for papillary thyroid carcinomas

INTRODUCTION

Papillary thyroid carcinoma (PTC) is the most prevalent form of thyroid cancer, with the classical and follicular variants representing most cases. Despite generally favorable prognoses, approximately 10% of patients experience recurrence post-surgery and radioactive iodine therapy. Attempts to stratify risk of recurrence have relied on gene expression-based prognostic and predictive signatures with a focus on mutations of well-known driver genes, while hallmarks of tumor morphology have been ignored.

OBJECTIVES

We introduce a new computational pathology approach to develop prognostic gene signatures for PTC that is informed by quantitative features of tumor and immune cell morphology.

METHODS

We quantified nuclear and immune-related features of tumor morphology to develop a pathomic signature, which was then used to inform an RNA-expression signature model provides a notable advancement in risk stratification compared to both standalone and pathology-informed gene-expression signatures.

RESULTS

There was a 17.8% improvement in the C-index (from 0.605 to 0.783) for 123 cPTCs and 15% (from 0.576 to 0.726) for 38 fvPTCs compared to the standalone gene-expression signature. Hazard ratios also improved for cPTCs from 0.89 (0.67,0.99) to 4.43 (3.65,6.68) and fvPTC from 0.98 (0.76,1.32) to 2.28 (1.87,3.64). We validated the image-based risk model on an independent cohort of 32 cPTCs with hazard ratio 1.8 (1.534,2.167).

Educational exchange in thyroid core needle biopsy diagnosis: enhancing pathological interpretation through guideline integration and peer learning

BACKGROUND

While fine needle aspiration cytology (FNAC) plays an essential role in the screening of thyroid nodules, core needle biopsy (CNB) acts as an alternative method to address FNAC limitations. However, diagnosing thyroid CNB samples can be challenging due to variations in background and levels of experience. Effective training is indispensable to mitigate this challenge. We aim to evaluate the impact of an educational program on improving the accuracy of CNB diagnostics.

METHODS

The 2-week observational program included a host mentor pathologist with extensive experience and a visiting pathologist. The CNB classification by The Practice Guidelines Committee of the Korean Thyroid Association was used for the report. Two rounds of reviewing the case were carried out, and the level of agreement between the reviewers was analyzed.

RESULTS

The first-round assessment showed a concordance between two pathologists for 247 thyroid CNB specimens by 84.2%, with a kappa coefficient of 0.74 (indicating substantial agreement). This finding was attributed to the discordance in the use of categories III and V. After peer learning, the two pathologists evaluated 30 new cases, which showed an overall improvement in the level of agreement. The percentage of agreement between pathologists on thyroid CNB diagnosis was 86.7%, as measured by kappa coefficient of 0.80.

CONCLUSIONS

This educational program, consisting of guided mentorship and peer learning, can substantially enhance the diagnostic accuracy of thyroid CNB. It is useful in promoting consistent diagnostic standards and contributes to the ongoing development of global pathology practices.

Differentiating BRAF V600E- and RAS-like alterations in encapsulated follicular patterned tumors through histologic features: a validation study

Differentiating BRAF V600E- and RAS-altered encapsulated follicular-patterned thyroid tumors based on morphology remains challenging. This study aimed to validate an 8-score scale nuclear scoring system and investigate the importance of nuclear pseudoinclusions (NPIs) in aiding this differentiation. A cohort of 44 encapsulated follicular-patterned tumors with varying degrees of nuclear atypia and confirmed BRAF V600E or RAS alterations was studied. Nuclear parameters (area, diameter, and optical density) were analyzed using a deep learning model. Twelve pathologists from eight Asian countries visually assessed 22 cases after excluding the cases with any papillae. Eight nuclear features were applied, yielding a semi-quantitative score from 0 to 24. A threshold score of 14 was used to distinguish between RAS- and BRAF V600E-altered tumors. BRAF V600E-altered tumors typically demonstrated higher nuclear scores and notable morphometric alterations. Specifically, the nuclear area and diameter were significantly larger, and nuclear optical density was much lower compared to RAS-altered tumors. Observer accuracy varied, with two pathologists correctly identifying genotype of all cases. Observers were categorized into proficiency groups, with the highest group maintaining consistent accuracy across both evaluation methods. The lower group showed a significant improvement in accuracy upon utilizing the 8-score scale nuclear scoring system, with notably increased sensitivity and negative predictive value in BRAF V600E tumor detection. BRAF V600E-altered tumors had higher median total nuclear scores. Detailed reevaluation revealed NPIs in all BRAF V600E-altered cases, but in only 2 of 14 RAS-altered cases. These results could significantly assist pathologists, particularly those not specializing in thyroid pathology, in making a more accurate diagnosis.

Deep Learning-Based Differential Diagnosis of Follicular Thyroid Tumors Using Histopathological Images

Deep learning systems (DLSs) have been developed for the histopathological assessment of various types of tumors, but none are suitable for differential diagnosis between follicular thyroid carcinoma (FTC) and follicular adenoma (FA). Furthermore, whether DLSs can identify the malignant characteristics of thyroid tumors based only on random views of tumor tissue histology has not been evaluated. In this study, we developed DLSs able to differentiate between FTC and FA based on 3 types of convolutional neural network architecture: EfficientNet, VGG16, and ResNet50. The performance of all 3 DLSs was excellent (area under the receiver operating characteristic curve = 0.91 ± 0.04; F1 score = 0.82 ± 0.06). Visual explanations using gradient-weighted class activation mapping suggested that the diagnosis of both FTC and FA was largely dependent on nuclear features. The DLSs were then trained with FTC images and linked information (presence or absence of recurrence within 10 years, vascular invasion, and wide capsular invasion). The ability of the DLSs to diagnose these characteristics was then determined. The results showed that, based on the random views of histology, the DLSs could predict the risk of FTC recurrence, vascular invasion, and wide capsular invasion with a certain level of accuracy (area under the receiver operating characteristic curve = 0.67 ± 0.13, 0.62 ± 0.11, and 0.65 ± 0.09, respectively). Further improvement of our DLSs could lead to the establishment of automated differential diagnosis systems requiring only biopsy specimens.

Preoperative Risk Stratification of Follicular-patterned Thyroid Lesions on Core Needle Biopsy by Histologic Subtyping and RAS Variant-specific Immunohistochemistry

Follicular-patterned lesions often have indeterminate results (diagnostic category III or IV) by core needle biopsy (CNB) and fine needle aspiration (FNA). However, CNB diagnoses follicular neoplasm (category IV) more frequently than FNA. Therefore, we aimed to develop a risk stratification system for CNB samples with category III/IV using immunohistochemistry (IHC). The specificity of the RAS Q61R antibody was validated on 58 thyroid nodules with six different types of RAS genetic variants and 40 cases of RAS wild-type. We then applied IHC analysis of RAS Q61R to 207 CNB samples with category III/IV in which all patients underwent surgical resection. RAS Q61R IHC had 98% sensitivity and 98% specificity for detecting the RAS p.Q16R variant. In an independent dataset, the positive rate of RAS Q61R was significantly higher in NIFTP (48%) and malignancies (45%) than in benign tumors (19%). The risk of NIFTP/malignancy was highest in the group with nuclear atypia and RAS Q61R expression (86%) and lowest in the group without both parameters (32%). The high-risk group with either nuclear atypia or RAS Q61R had 67.3% sensitivity, 73.4% specificity, 75.2% positive predictive value, and 65.1% negative predictive value for identifying NIFTP/malignancy. We conclude that RAS Q61R IHC can be a rule-in diagnostic test for NIFTP/malignancy in CNB category III/IV results. Combining of the histologic parameter (nuclear atypia) with RAS Q61R IHC results can further stratify CNB category III/IV into a high-risk group, which is sufficient for a surgical referral, and a low-risk group sufficient for observation.