aetherAI—Asia’s leading medical image AI solution provider focused on digital pathology and medical imaging AI—has announced the publication of its machine learning approach for the accurate quantification of nuclear morphometrics and differential diagnosis of primary intestinal T-cell lymphomas in the peer-reviewed Cancers. This approach not only brings deeper insight into lymphoma phenotypes but also paves the way for future discoveries concerning their relationship with disease classification and outcomes. Entitled “Machine Learning Based on Morphological Features Enables Classification of Primary Intestinal T-Cell Lymphomas,” the article can be accessed online here.
Fig.1: Sample images used as the dataset.
(A) An example of the whole-slide image. A high-quality region was manually marked with a black circle by a senior hematopathologist.
(B) Left to right: Representative HPF images of neoplastic lymphocytes from MEITL, borderline, and ITCL-NOS cases. The neoplastic lymphocytes detected and segmented by the model are contoured in red.
The aim of the study was to investigate the feasibility of using machine learning techniques based on morphological features to classify two subtypes of primary intestinal T-cell lymphomas (PITLs): monomorphic epitheliotropic intestinal T-cell lymphoma (MEITL) and intestinal T-cell lymphoma, not otherwise specified (ITCL-NOS), defined according to WHO criteria. These features are considered a major challenge for pathological diagnosis.
The results indicate that the XGBoost model gave superior classification performance to the end-to-end CNN model and could elucidate explicit relationships between predictions and morphological features, while also achieving results comparable to classifications incorporating immunophenotype or to work by senior hematopathologists.
The deep neural network achieved an average precision of 0.881 in the cell segmentation work. In its classification of MEITL versus ITCL-NOS, the XGBoost model achieved an area under the receiver operating characteristic curve (AUC) of 0.966.
Fig. 2: Comparison of contour segmentation results with the QuPath cell detection module. aetherAI’s lymphocyte detection model (third row) showed fewer FC and FP but more false negatives (FN) at a score threshold of 0.2.
“Our research demonstrates an accurate, human-interpretable approach to using machine learning algorithms to reduce the high dimensionality of image features and classify T cell lymphomas that present challenges in morphologic diagnosis,” said Joe Yeh, M.D., aetherAI CEO. The quantitative nuclear morphometric features may lead to further discoveries concerning the relationship between cellular phenotype and disease status. “Our model may have great potential to improve diagnostic consistency, efficiency, and accuracy for other types of cancers,” Yeh added.
The study was conducted in collaboration with Dr. Shih-Sung Chuang of the Department of Pathology at Chi-Mei Medical Center in Taiwan and with four other hospital departments of pathology. A total of 40 histopathological whole-slide images (WSIs) from 40 surgically resected PITL cases were used as the dataset for model training and testing.
運用AI輔助診斷罕見腫瘤 實現精準醫療
雲象科技共同研究登上國際權威期刊
數位病理及醫療影像AI領導廠商 - 雲象科技所研發的醫療影像AI應用，成功協助醫師對淋巴瘤可有更細緻且精確的診斷方法，這項研究成果已獲刊登於國際知名醫療期刊《Cancers》，肯定技術應用的突破。
T細胞淋巴瘤是少見的疾病，最重要的第一步是正確的病理診斷，精確診斷則仰賴病理科醫師對於腫瘤組織切片的形態分析。T細胞淋巴瘤有兩類在形態上難以區分的疾病：1、單形性上皮腸T細胞淋巴瘤 (MEITL，monomorphic epitheliotropic intestinal T-cell lymphoma)；2、腸道T細胞淋巴瘤 (ITCL-NOS，intestinal T-cell lymphoma, not otherwise specified)。病理診斷是透過細胞核大小變異度、形態分布、以及免疫表現型來區分這兩類疾病，對經驗尚在累積的一般病理科醫師挑戰甚鉅，而且形態介於兩個極端之間的案例也難以確切診斷。
以機器學習訓練細胞核形態分析
奇美醫院醫學中心病理部部長兼解剖病理科主任莊世松教授與雲象科技合作，運用數千顆細胞核的標註資料，訓練出的AI模型能對淋巴瘤的細胞核精確地偵測，描繪輪廓，並且進一步計算出每個細胞核面積大小，長短軸比例等可量化的形態資訊，依此進一步訓練機器學習演算法，對兩種類型MEITL、ITCL-NOS進行分類，預測水準可高達AUC 0.966（完美演算法的AUC為1）。
免除粗糙二元分類，精準AI量化分析輔助
透過如此AI輔助診斷工具，可以讓醫師不再需要對T細胞淋巴瘤的形態進行粗糙的二元分類，而可以用量化的形態分析數據，對於淋巴瘤細胞作精確的統計描述。雲象科技執行長葉肇元醫師表示，相信以此AI量化分析，可以讓淋巴瘤有更細緻、精準的診斷方法，醫師因此可以施以更精確的治療。雲象持續精進科學研究，這次共同發表是計算病理學(computational pathology)能力的初步展現，更多的發揮空間指日可待，將持續提升診斷及治療的水準。
News coverage:
Comentarios