Clinical Interpretability of Deep Learning for Predicting Microvascular Invasion in Hepatocellular Carcinoma by Using Attention Mechanism
Preoperative prediction of microvascular invasion (MVI) is essential for management decision in hepatocellular carcinoma (HCC). Deep learning-based prediction models of MVI are numerous but lack clinical interpretation due to their “black-box” nature. Consequently, we aimed to use an attention-guide...
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-08-01
|
| Series: | Bioengineering |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2306-5354/10/8/948 |
| _version_ | 1797585431452188672 |
|---|---|
| author | Huayu You Jifei Wang Ruixia Ma Yuying Chen Lujie Li Chenyu Song Zhi Dong Shiting Feng Xiaoqi Zhou |
| author_facet | Huayu You Jifei Wang Ruixia Ma Yuying Chen Lujie Li Chenyu Song Zhi Dong Shiting Feng Xiaoqi Zhou |
| author_sort | Huayu You |
| collection | DOAJ |
| description | Preoperative prediction of microvascular invasion (MVI) is essential for management decision in hepatocellular carcinoma (HCC). Deep learning-based prediction models of MVI are numerous but lack clinical interpretation due to their “black-box” nature. Consequently, we aimed to use an attention-guided feature fusion network, including intra- and inter-attention modules, to solve this problem. This retrospective study recruited 210 HCC patients who underwent gadoxetate-enhanced MRI examination before surgery. The MRIs on pre-contrast, arterial, portal, and hepatobiliary phases (hepatobiliary phase: HBP) were used to develop single-phase and multi-phase models. Attention weights provided by attention modules were used to obtain visual explanations of predictive decisions. The four-phase fusion model achieved the highest area under the curve (AUC) of 0.92 (95% CI: 0.84–1.00), and the other models proposed AUCs of 0.75–0.91. Attention heatmaps of collaborative-attention layers revealed that tumor margins in all phases and peritumoral areas in the arterial phase and HBP were salient regions for MVI prediction. Heatmaps of weights in fully connected layers showed that the HBP contributed the most to MVI prediction. Our study firstly implemented self-attention and collaborative-attention to reveal the relationship between deep features and MVI, improving the clinical interpretation of prediction models. The clinical interpretability offers radiologists and clinicians more confidence to apply deep learning models in clinical practice, helping HCC patients formulate personalized therapies. |
| first_indexed | 2024-03-11T00:06:00Z |
| format | Article |
| id | doaj.art-06aa9213cf994722a81d53b71509c0d8 |
| institution | Directory Open Access Journal |
| issn | 2306-5354 |
| language | English |
| last_indexed | 2024-03-11T00:06:00Z |
| publishDate | 2023-08-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Bioengineering |
| spelling | doaj.art-06aa9213cf994722a81d53b71509c0d82023-11-19T00:18:14ZengMDPI AGBioengineering2306-53542023-08-0110894810.3390/bioengineering10080948Clinical Interpretability of Deep Learning for Predicting Microvascular Invasion in Hepatocellular Carcinoma by Using Attention MechanismHuayu You0Jifei Wang1Ruixia Ma2Yuying Chen3Lujie Li4Chenyu Song5Zhi Dong6Shiting Feng7Xiaoqi Zhou8Department of Radiology, The First Affiliated Hospital, Sun Yat-sen University, 58th the Second Zhongshan Road, Guangzhou 510080, ChinaDepartment of Radiology, The First Affiliated Hospital, Sun Yat-sen University, 58th the Second Zhongshan Road, Guangzhou 510080, ChinaDepartment of Radiology, The First Affiliated Hospital, Sun Yat-sen University, 58th the Second Zhongshan Road, Guangzhou 510080, ChinaDepartment of Radiology, The First Affiliated Hospital, Sun Yat-sen University, 58th the Second Zhongshan Road, Guangzhou 510080, ChinaDepartment of Radiology, The First Affiliated Hospital, Sun Yat-sen University, 58th the Second Zhongshan Road, Guangzhou 510080, ChinaDepartment of Radiology, The First Affiliated Hospital, Sun Yat-sen University, 58th the Second Zhongshan Road, Guangzhou 510080, ChinaDepartment of Radiology, The First Affiliated Hospital, Sun Yat-sen University, 58th the Second Zhongshan Road, Guangzhou 510080, ChinaDepartment of Radiology, The First Affiliated Hospital, Sun Yat-sen University, 58th the Second Zhongshan Road, Guangzhou 510080, ChinaDepartment of Radiology, The First Affiliated Hospital, Sun Yat-sen University, 58th the Second Zhongshan Road, Guangzhou 510080, ChinaPreoperative prediction of microvascular invasion (MVI) is essential for management decision in hepatocellular carcinoma (HCC). Deep learning-based prediction models of MVI are numerous but lack clinical interpretation due to their “black-box” nature. Consequently, we aimed to use an attention-guided feature fusion network, including intra- and inter-attention modules, to solve this problem. This retrospective study recruited 210 HCC patients who underwent gadoxetate-enhanced MRI examination before surgery. The MRIs on pre-contrast, arterial, portal, and hepatobiliary phases (hepatobiliary phase: HBP) were used to develop single-phase and multi-phase models. Attention weights provided by attention modules were used to obtain visual explanations of predictive decisions. The four-phase fusion model achieved the highest area under the curve (AUC) of 0.92 (95% CI: 0.84–1.00), and the other models proposed AUCs of 0.75–0.91. Attention heatmaps of collaborative-attention layers revealed that tumor margins in all phases and peritumoral areas in the arterial phase and HBP were salient regions for MVI prediction. Heatmaps of weights in fully connected layers showed that the HBP contributed the most to MVI prediction. Our study firstly implemented self-attention and collaborative-attention to reveal the relationship between deep features and MVI, improving the clinical interpretation of prediction models. The clinical interpretability offers radiologists and clinicians more confidence to apply deep learning models in clinical practice, helping HCC patients formulate personalized therapies.https://www.mdpi.com/2306-5354/10/8/948hepatocellular carcinomamicrovascular invasionattention mechanismdeep learningmulti-phase MRI |
| spellingShingle | Huayu You Jifei Wang Ruixia Ma Yuying Chen Lujie Li Chenyu Song Zhi Dong Shiting Feng Xiaoqi Zhou Clinical Interpretability of Deep Learning for Predicting Microvascular Invasion in Hepatocellular Carcinoma by Using Attention Mechanism Bioengineering hepatocellular carcinoma microvascular invasion attention mechanism deep learning multi-phase MRI |
| title | Clinical Interpretability of Deep Learning for Predicting Microvascular Invasion in Hepatocellular Carcinoma by Using Attention Mechanism |
| title_full | Clinical Interpretability of Deep Learning for Predicting Microvascular Invasion in Hepatocellular Carcinoma by Using Attention Mechanism |
| title_fullStr | Clinical Interpretability of Deep Learning for Predicting Microvascular Invasion in Hepatocellular Carcinoma by Using Attention Mechanism |
| title_full_unstemmed | Clinical Interpretability of Deep Learning for Predicting Microvascular Invasion in Hepatocellular Carcinoma by Using Attention Mechanism |
| title_short | Clinical Interpretability of Deep Learning for Predicting Microvascular Invasion in Hepatocellular Carcinoma by Using Attention Mechanism |
| title_sort | clinical interpretability of deep learning for predicting microvascular invasion in hepatocellular carcinoma by using attention mechanism |
| topic | hepatocellular carcinoma microvascular invasion attention mechanism deep learning multi-phase MRI |
| url | https://www.mdpi.com/2306-5354/10/8/948 |
| work_keys_str_mv | AT huayuyou clinicalinterpretabilityofdeeplearningforpredictingmicrovascularinvasioninhepatocellularcarcinomabyusingattentionmechanism AT jifeiwang clinicalinterpretabilityofdeeplearningforpredictingmicrovascularinvasioninhepatocellularcarcinomabyusingattentionmechanism AT ruixiama clinicalinterpretabilityofdeeplearningforpredictingmicrovascularinvasioninhepatocellularcarcinomabyusingattentionmechanism AT yuyingchen clinicalinterpretabilityofdeeplearningforpredictingmicrovascularinvasioninhepatocellularcarcinomabyusingattentionmechanism AT lujieli clinicalinterpretabilityofdeeplearningforpredictingmicrovascularinvasioninhepatocellularcarcinomabyusingattentionmechanism AT chenyusong clinicalinterpretabilityofdeeplearningforpredictingmicrovascularinvasioninhepatocellularcarcinomabyusingattentionmechanism AT zhidong clinicalinterpretabilityofdeeplearningforpredictingmicrovascularinvasioninhepatocellularcarcinomabyusingattentionmechanism AT shitingfeng clinicalinterpretabilityofdeeplearningforpredictingmicrovascularinvasioninhepatocellularcarcinomabyusingattentionmechanism AT xiaoqizhou clinicalinterpretabilityofdeeplearningforpredictingmicrovascularinvasioninhepatocellularcarcinomabyusingattentionmechanism |