Use of steady-state thermography and dynamic thermography in breast cancer diagnosis

背景與目的： 乳房熱攝影是一偵測腫瘤之生理現象之檢查。我們的研究目的，是欲評估兩類熱攝影- 靜態及動態熱攝影之診斷表現，並要分析乳房熱攝影表徵與預測乳癌預後因子之相關性。這些預測乳癌預後因子包括動情激素接受體 (ER) 及黃體素接受體 (PR)、人體上皮生長因子接受體2 (HER2)、臨床分級、組織分級。 方法： 本研究為橫斷性研究，是以在乳房X光攝影或超音波有局部病灶之女性為研究對象。我們將由五種熱攝影表徵樹立靜態熱攝影評分系統。我們將把年齡及上述表徵置入模型，並以多變數邏輯式迴歸模型及接受操作特徵曲線 (Receiver operating characteristics, ROC) 來評估診斷表現。 在接受靜態及動態熱攝影之次族群，判讀表徵包括二靜態及二動態表徵。我們將由只含靜態表徵 (模型A) 及含靜態、動態表徵之年齡調整模型 (模型B)，估計二種模型之接受操作特徵曲線下面積 (Area under the ROC curve, AUC)。 我們將分析靜態及動態熱攝影表徵與乳癌預後因子相關性之評估。 結果： 在靜態熱攝影方面，共有298個病灶接受分析。其適切模型之AUC為0.781 (95%信賴區間0.683-0.855)。當敏感度為94.3%時，特異度為30.4%。在171個惡性病灶中，二表徵與ER成反向相關 (P值0.010，0.037)。三表徵和PR成反向相關 (P值0.039，0.020，0.022)。Triple-negative (即ER、PR、HER2皆陰性) 之乳癌傾向有較高之熱攝影評分 (和其他類相比，P值0.029)。一表徵和臨床分級正相關 (P值0.008)。在128例侵襲性乳管癌中，一表徵和組織分級成正相關 (P值0.037)。 本研究中有56個病灶有靜態及動態熱攝影資料供分析。由模型A及B 之 AUC均為0.744 (95%信賴區間: 模型A，0.512-0.978; B，0.514-0.975)。在敏感度100%時，二模型之特異度均為33.3%。在其中之26惡性病灶中，一動態表徵和ER成反向相關 (P值0.021)，二動態表徵和HER2成正相關 (P值0.015，0.033)。一靜態表徵與侵襲性乳管癌組織分級成正相關 (P值0.006，22例病灶)。 結論： 當乳房熱攝影之敏感度>90%時，特異度相對地偏低。動態熱攝影並未比靜態熱攝影更顯著地增加診斷準確度。乳房熱攝影表徵可能在將來可以做為預測乳癌預後之指標。Background and Objective: Breast thermography is an examination which delineates physiological response of the tumor. The purpose of our study is to evaluate the diagnostic performance of the two types of breast thermography- steady-state and dynamic thermography, and investigate the association of thermographic signs with breast cancer prognostic indicators, including ER (estrogen receptor), PR (progesterone receptor), HER2 (human epidermal growth factor receptor 2) statuses, clinical staging, and histologic grade. Methods: We conducted a cross-sectional study, and enrolled women who had suspicious findings on mammography or breast ultrasound. We interpreted steady-state thermography using five thermographic signs and evaluated the diagnostic performance by ROC analysis (Receiver operating characteristics) from age-adjusted multiple logistic regression model For the subpopulation of women receiving both steady-state and dynamic thermography, the interpretation was determined by the two steady-state signs and two dynamic signs. The age-adjusted logistic regression models with only steady-state signs (model A) and with both steady-state and dynamic signs (model B), and the resulting AUC (Area under the ROC curve) can be derived. The association of thermographic signs and the prognostic indicators of the cancerous lesions were estimated for steady-state thermography and dynamic thermography. Results: A total of 298 breast lesions were included for steady-state thermographic study. The AUC was 0.781 for the fitted model (95% CI: 0.683-0.855). The specificity was 30.4% when the sensitivity was 94.3%. Of the 171 cancerous lesions, two thermographic signs were inversely associated with ER (P=0.010 and 0.037), and three signs were inversely associated with PR (P=0.039, 0.020, and 0.022). Triple-negative (ER, PR, and HER2 negative) cancers tended to show higher thermographic scores than other types of cancers (P =0.029). One thermographic sign was positively associated with clinical staging (P=0.008). One thermographic sign was positively associated with histologic grade of invasive ductal carcinoma (n=128; P=0.037). There were 56 lesions with both steady-state and dynamic images available. The AUC was both 0.744 for the models A and B (95% CI: model A, 0.512-0.978; B, 0.514-0.975), and the specificity was 33.3% when the sensitivity was 100% for both models. Of the 26 malignant lesions, one of the dynamic signs was inversely associated with ER (P=0.021). The two dynamic signs were positively associated with HER2 status (P=0.015 and 0.033). One steady-state sign was positively related with the histologic grade of invasive ductal carcinoma (P=0.006; n=22). Conclusion: The specificity of breast thermography was low when the sensitivity >90%. Dynamic thermography did not show additional diagnostic yield than steady-state thermography. The thermographic signs may be predictive of breast cancer prognosis