神山 直久¹, 杉本 勝俊², 中原 龍一³, 掛川 達矢², 糸井 隆夫²

Naohisa KAMIYAMA¹, Katsutoshi SUGIMOTO², Ryuichi NAKAHARA³, Tatsuya KAKEGAWA², Takao ITOI²

¹GEヘルスケア・ジャパン超音波製品開発部, ²東京医科大学消化器内科, ³岡山大学病院整形外科

¹Ultrasound General Imaging, GE HealthCare Japan, ²Department of Gastroenterology and Hepatology, Tokyo Medical University, ³Department of Orthopedic Surgery, Dentistry and Pharmaceutical Sciences, Okayama University Graduate School of Medicine

キーワード : hepatocellular carcinoma, contrast-enhanced ultrasonography, machine learning, multi-input deep learning model

目的：造影超音波検査（contrast-enhanced ultrasound：CEUS）では，造影剤投与後の経過時間によって造影パターンが変化する．本研究では，9時相のCEUS画像を入力とする独自の深層学習モデルを用いて，肝結節の性状診断能を評価することを目的とした．方法：前向き試験として，肝病変181例〔良性48例，肝細胞癌（hepatocellular carcinoma：HCC）78例，非HCC悪性腫瘍55例〕を対象とした．造影剤にはSonazoidを用い，注入前のBモード画像に加え，投与後10分まで1分ごとに画像クリップを記録した．深層学習モデルは3つのResNet50転移学習モデルを並列に配置して構築した．本モデルではCEUSの最大9相までのデータセットを同時入力でき，9枚の画像を同期して画像拡張処理することが可能である．結果から「良性」と「悪性」の正診率，感度，特異度を時相の組み合わせごとに算出し，これらの精度指標を読影医が判定したウォッシュアウトスコアと比較した．結果：提案モデルは，Bモードから投与後10分までの画像を含むデータセットを用いた場合に参照標準モデルを上回る性能を示した（感度93.2％，特異度65.3％，平均正診率60.1％）．また，データセットを造影剤投与後2分以内に限定しても感度90.2％，特異度61.2％を維持し，この精度は専門医によるウォッシュアウトスコア判定と同等以上であった．結論：本モデルは，クッパー相より早い段階で腫瘍の種類を判別できる可能性を示した．同時に，機械学習の結果から，Sonazoidを用いたクッパー相画像には肝結節を分類するうえで不可欠な情報が含まれていることが確認された．

Purpose: Contrast-enhanced ultrasound（CEUS） shows different enhancement patterns depending on the time after administration of the contrast agent. The aim of this study was to evaluate the diagnostic performance of liver nodule characterization using our proposed deep learning model with input of nine CEUS images. Methods: A total of 181 liver lesions （48 benign, 78 hepatocellular carcinoma（HCC）, and 55 non-HCC malignant） were included in this prospective study. CEUS were performed using the contrast agent Sonazoid, and in addition to B-mode images before injection, image clips were stored every minute up to 10 min. A deep learning model was developed by arranging three ResNet50 transfer learning models in parallel. This proposed model allowed inputting up to nine datasets of different phases of CEUS and performing image augmentation of nine images synchronously. Using the results, the correct prediction rate, sensitivity, and specificity between “benign” and “malignant” cases were analyzed for each combination of the time phase. These accuracy values were also compared with the washout score judged by a human. Results: The proposed model showed performance superior to the referential standard model when the dataset from B-mode to the 10-min images were used（sensitivity:93.2％, specificity: 65.3％, average correct answer rate:60.1％）. It also maintained 90.2％ sensitivity and 61.2％ specificity even when the dataset was limited to 2 min after injection, and this accuracy was equivalent to or better than human scoring by experts. Conclusion: Our proposed model has the potential to identify tumor types earlier than the Kupffer phase, but at the same time, machine learning confirmed that Kupffer-phase Sonazoid images contain essential information for the classification of liver nodules.