Bio


I pursue my career path as a physician-scientist in the field of biomedical informatics. My focus is translational research aiming to develop and incorporate AI into clinical workflow. I have conducted research on artificial intelligence (AI) system to improve ultrasound surveillance for hepatocellular carcinoma (HCC), machine learning models for predicting HCC development in high-risk patients, Ultrasound Liver Imaging Reporting and Data System (US LI-RADS) for B-mode and contrast-enhanced ultrasound, as well as AI techniques for quantitative analysis of such imaging modalities.

Education & Certifications


  • PhD, Chulalongkorn University, Bangkok, Thailand, Clinical Sciences (International Program) (2023)
  • MD (First class honors), Chulalongkorn University, Bangkok, Thailand, Doctor of Medicine (2020)

Current Research and Scholarly Interests


artificial intelligence
medical imaging
ultrasound
screening and surveillance of liver cancer
cancer prediction models
cancer biomarkers

All Publications


  • Artificial intelligence assists operators in real-time detection of focal liver lesions during ultrasound: A randomized controlled study EUROPEAN JOURNAL OF RADIOLOGY Tiyarattanachai, T., Apiparakoon, T., Chaichuen, O., Sukcharoen, S., Yimsawad, S., Jangsirikul, S., Chaikajornwat, J., Siriwong, N., Burana, C., Siritaweechai, N., Atipas, K., Assawamasbunlue, N., Tovichayathamrong, P., Obcheuythed, P., Somvanapanich, P., Geratikornsupuk, N., Anukulkarnkusol, N., Sarakul, P., Tanpowpong, N., Pinjaroen, N., Kerr, S. J., Rerknimitr, R., Marukatat, S., Chaiteerakij, R. 2023; 165: 110932

    Abstract

    Detection of hepatocellular carcinoma (HCC) is crucial during surveillance by ultrasound. We previously developed an artificial intelligence (AI) system based on convolutional neural network for detection of focal liver lesions (FLLs) in ultrasound. The primary aim of this study was to evaluate whether the AI system can assist non-expert operators to detect FLLs in real-time, during ultrasound examinations.This single-center prospective randomized controlled study evaluated the AI system in assisting non-expert and expert operators. Patients with and without FLLs were enrolled and had ultrasound performed twice, with and without AI assistance. McNemar's test was used to compare paired FLL detection rates and false positives between groups with and without AI assistance.260 patients with 271 FLLs and 244 patients with 240 FLLs were enrolled into the groups of non-expert and expert operators, respectively. In non-experts, FLL detection rate in the AI assistance group was significantly higher than the no AI assistance group (36.9 % vs 21.4 %, p < 0.001). In experts, FLL detection rates were not significantly different between the groups with and without AI assistance (66.7 % vs 63.3 %, p = 0.32). False positive detection rates in the groups with and without AI assistance were not significantly different in both non-experts (14.2 % vs 9.2 %, p = 0.08) and experts (8.6 % vs 9.0 %, p = 0.85).The AI system resulted in significant increase in detection of FLLs during ultrasound examinations by non-experts. Our findings may support future use of the AI system in resource-limited settings where ultrasound examinations are performed by non-experts. The study protocol was registered under the Thai Clinical Trial Registry (TCTR20201230003), which is part of the WHO ICTRP Registry Network. The registry can be accessed via the following URL: https://trialsearch.who.int/Trial2.aspx?TrialID=TCTR20201230003.

    View details for DOI 10.1016/j.ejrad.2023.110932

    View details for Web of Science ID 001027781300001

    View details for PubMedID 37390663

  • A Comprehensive Motion Compensation Method for In-Plane and Out-of-Plane Motion in Dynamic Contrast-Enhanced Ultrasound of Focal Liver Lesions. Ultrasound in medicine & biology Tiyarattanachai, T., Turco, S., Eisenbrey, J. R., Wessner, C. E., Medellin-Kowalewski, A., Wilson, S., Lyshchik, A., Kamaya, A., Kaffas, A. E. 2022

    Abstract

    Contrast-enhanced ultrasound (CEUS) acquisitions of focal liver lesions are affected by motion, which has an impact on contrast signal quantification. We therefore developed and tested, in a large patient cohort, a motion compensation algorithm called the Iterative Local Search Algorithm (ILSA), which can correct for both periodic and non-periodic in-plane motion and can reject frames with out-of-plane motion. CEUS cines of 183 focal liver lesions in 155 patients from three hospitals were used to develop and test ILSA. Performance was evaluated through quantitative metrics, including the root mean square error and R2 in fitting time-intensity curves and standard deviation value of B-mode intensities, computed across cine frames), and qualitative evaluation, including B-mode mean intensity projection images and parametric perfusion imaging. The median root mean square error significantly decreased from 0.032 to 0.024 (p < 0.001). Median R2 significantly increased from 0.88 to 0.93 (p < 0.001). The median standard deviation value of B-mode intensities significantly decreased from 6.2 to 5.0 (p < 0.001). B-Mode mean intensity projection images revealed improved spatial resolution. Parametric perfusion imaging also exhibited improved spatial detail and better differentiation between lesion and background liver parenchyma. ILSA can compensate for all types of motion encountered during liver CEUS, potentially improving contrast signal quantification of focal liver lesions.

    View details for DOI 10.1016/j.ultrasmedbio.2022.06.007

    View details for PubMedID 35970658

  • The feasibility to use artificial intelligence to aid detecting focal liver lesions in real-time ultrasound: a preliminary study based on videos SCIENTIFIC REPORTS Tiyarattanachai, T., Apiparakoon, T., Marukatat, S., Sukcharoen, S., Yimsawad, S., Chaichuen, O., Bhumiwat, S., Tanpowpong, N., Pinjaroen, N., Rerknimitr, R., Chaiteerakij, R. 2022; 12 (1): 7749

    Abstract

    Despite the wide availability of ultrasound machines for hepatocellular carcinoma surveillance, an inadequate number of expert radiologists performing ultrasounds in remote areas remains a primary barrier for surveillance. We demonstrated feasibility of artificial intelligence (AI) to aid in the detection of focal liver lesions (FLLs) during ultrasound. An AI system for FLL detection in ultrasound videos was developed. Data in this study were prospectively collected at a university hospital. We applied a two-step training strategy for developing the AI system by using a large collection of ultrasound snapshot images and frames from full-length ultrasound videos. Detection performance of the AI system was evaluated and then compared to detection performance by 25 physicians including 16 non-radiologist physicians and 9 radiologists. Our dataset contained 446 videos (273 videos with 387 FLLs and 173 videos without FLLs) from 334 patients. The videos yielded 172,035 frames with FLLs and 1,427,595 frames without FLLs for training on the AI system. The AI system achieved an overall detection rate of 89.8% (95%CI: 84.5-95.0) which was significantly higher than that achieved by non-radiologist physicians (29.1%, 95%CI: 21.2-37.0, p < 0.001) and radiologists (70.9%, 95%CI: 63.0-78.8, p < 0.001). Median false positive detection rate by the AI system was 0.7% (IQR: 1.3%). AI system operation speed reached 30-34 frames per second, showing real-time feasibility. A further study to demonstrate whether the AI system can assist operators during ultrasound examinations is warranted.

    View details for DOI 10.1038/s41598-022-11506-z

    View details for Web of Science ID 000794011500033

    View details for PubMedID 35545628

    View details for PubMedCentralID PMC9095624

  • Multicenter Study of ACR Ultrasound LI-RADS Visualization Scores on Serial Examinations: Implications for Changes in Surveillance Strategies. AJR. American journal of roentgenology Tiyarattanachai, T., Fetzer, D. T., Kamaya, A. 2022

    Abstract

    Background: American College of Radiology Ultrasound LI-RADS includes the visualization score as a subjective measure of examination quality and expected level of sensitivity. Whether a single suboptimal visualization score warrants change in surveillance strategy is unknown. Objective: To determine the relative stability of visualization scores on serial surveillance ultrasound examinations in patients at risk for HCC. Methods: This retrospective study included patients at risk for HCC who underwent at least two HCC surveillance ultrasound examinations at one of three institutions between January 2017 and November 2020. Frequencies of score remaining unchanged after variable numbers of preceding examinations with the given score were determined. A mixed-effects logistic model was fitted to identify factors associated with a repeat score C (severe limitations) versus change to score A (no or minimal limitations) or score B (moderate limitations). Results: A total of 3169 patients underwent at least 2 ultrasound examinations, yielding a total of 9602 examinations. A total of 8030 (83.6%) examinations had score A, 1378 (14.4%) score B, and 194 (2.0%) score C. Frequency of score A was 88%, 91%, and 93% after 1, 2 and 3 consecutive prior examinations with score A. Frequency of score B was 45%, 48%, and 55% after 1, 2, and 3 consecutive prior examinations with score B. Frequency of score C was 42%, 67%, and 80% after 1, 2, and 3 consecutive prior examinations with score C. Among 109 examinations with score C in 91 patients with an available follow-up examination, no factor (including age, sex, severe steatosis, advanced cirrhosis, ascites, body mass index, and change in ultrasound machine, sonographer, or radiologist) was significantly associated with repeat score C (all p>.05). Although not statistically significant, presence of severe steatosis and advanced cirrhosis had the highest odds ratios (2.88 and 2.38, respectively) for repeat score C in multivariable analysis. Conclusion: Only 42% of patients with visualization score C on surveillances examination have score C on follow-up examination. Clinical Impact: The findings may inform decisions for alternative surveillance strategies in patients with visualization score C on ultrasound. This decision should consider the number of previous examinations with score C.

    View details for DOI 10.2214/AJR.22.27405

    View details for PubMedID 35383486

  • Ultrasound Liver Imaging Reporting and Data System (US LI-RADS) Visualization Score: a reliability analysis on inter-reader agreement. Abdominal radiology (New York) Tiyarattanachai, T., Bird, K. N., Lo, E. C., Mariano, A. T., Ho, A. A., Ferguson, C. W., Chima, R. S., Desser, T. S., Morimoto, L. N., Kamaya, A. 2021

    Abstract

    BACKGROUND & AIM: The American College of Radiology Ultrasound Liver Imaging Reporting and Data System (ACR US LI-RADS) Visualization Score conveys the expected level of sensitivity of screening and surveillance ultrasound exams in patients at risk for hepatocellular carcinoma (HCC). We sought to determine inter-reader agreement of the Visualization Score which is currently unknown.METHODS: Consecutive 6998 ultrasound HCC screening and surveillance studies in 3115 patients from 2017 to 2020 were retrospectively retrieved. Of these, 6154 (87.9%) studies were Visualization A (No or minimal limitations), 709 (10.1%) were Visualization B (Moderate limitations), and 135 (1.9%) were Visualization C (Severe limitations). Randomly sampled 90 studies, with 30 studies in each Visualization category, were included for analysis. Nine radiologists (3 senior attendings, 3 junior attendings and 3 body imaging fellows) blinded to the original categorization independently reviewed each study and assigned a Visualization Score. Intraclass correlation coefficient (ICC) was used to quantify inter-reader agreement.RESULTS: ICC among all 9 radiologists was 0.70 (95% CI 0.63-0.77). ICCs among senior attendings, junior attendings and body imaging fellows were 0.68 (CI 0.58-0.76), 0.72 (CI 0.62-0.80) and 0.76 (CI 0.68-0.83), respectively. Subgroup analysis by liver parenchyma was further performed. ICC was highest in the patient group with normal liver parenchyma (0.69, CI 0.56-0.81), followed by steatosis (0.66, CI 0.54-0.79) and cirrhosis (0.58, CI 0.43-0.73), respectively.CONCLUSIONS: US LI-RADS Visualization Score is a reliable tool with good inter-reader agreement that can be used to indicate the expected level of sensitivity of a screening and surveillance ultrasound examination for detecting focal liver observations.

    View details for DOI 10.1007/s00261-021-03067-y

    View details for PubMedID 34228197

  • Development and validation of artificial intelligence to detect and diagnose liver lesions from ultrasound images PLOS ONE Tiyarattanachai, T., Apiparakoon, T., Marukatat, S., Sukcharoen, S., Geratikornsupuk, N., Anukulkarnkusol, N., Mekaroonkamol, P., Tanpowpong, N., Sarakul, P., Rerknimitr, R., Chaiteerakij, R. 2021; 16 (6): e0252882

    Abstract

    Artificial intelligence (AI) using a convolutional neural network (CNN) has demonstrated promising performance in radiological analysis. We aimed to develop and validate a CNN for the detection and diagnosis of focal liver lesions (FLLs) from ultrasonography (USG) still images. The CNN was developed with a supervised training method using 40,397 retrospectively collected images from 3,487 patients, including 20,432 FLLs (hepatocellular carcinomas (HCCs), cysts, hemangiomas, focal fatty sparing, and focal fatty infiltration). AI performance was evaluated using an internal test set of 6,191 images with 845 FLLs, then externally validated using 18,922 images with 1,195 FLLs from two additional hospitals. The internal evaluation yielded an overall detection rate, diagnostic sensitivity and specificity of 87.0% (95%CI: 84.3-89.6), 83.9% (95%CI: 80.3-87.4), and 97.1% (95%CI: 96.5-97.7), respectively. The CNN also performed consistently well on external validation cohorts, with a detection rate, diagnostic sensitivity and specificity of 75.0% (95%CI: 71.7-78.3), 84.9% (95%CI: 81.6-88.2), and 97.1% (95%CI: 96.5-97.6), respectively. For diagnosis of HCC, the CNN yielded sensitivity, specificity, and negative predictive value (NPV) of 73.6% (95%CI: 64.3-82.8), 97.8% (95%CI: 96.7-98.9), and 96.5% (95%CI: 95.0-97.9) on the internal test set; and 81.5% (95%CI: 74.2-88.8), 94.4% (95%CI: 92.8-96.0), and 97.4% (95%CI: 96.2-98.5) on the external validation set, respectively. CNN detected and diagnosed common FLLs in USG images with excellent specificity and NPV for HCC. Further development of an AI system for real-time detection and characterization of FLLs in USG is warranted.

    View details for DOI 10.1371/journal.pone.0252882

    View details for Web of Science ID 000664641500027

    View details for PubMedID 34101764

    View details for PubMedCentralID PMC8186767

  • VOCs from Exhaled Breath for the Diagnosis of Hepatocellular Carcinoma DIAGNOSTICS Sukaram, T., Apiparakoon, T., Tiyarattanachai, T., Ariyaskul, D., Kulkraisri, K., Marukatat, S., Rerknimitr, R., Chaiteerakij, R. 2023; 13 (2)

    Abstract

    Volatile organic compound (VOC) profiles as biomarkers for hepatocellular carcinoma (HCC) are understudied. We aimed to identify VOCs from the exhaled breath for HCC diagnosis and compare the performance of VOCs to alpha-fetoprotein (AFP). The performance of VOCs for predicting treatment response and the association between VOCs level and survival of HCC patients were also determined.VOCs from 124 HCC patients and 219 controls were identified using the XGBoost algorithm. ROC analysis was used to determine VOCs performance in differentiating HCC patients from controls and in discriminating treatment responders from non-responders. The association between VOCs and the survival of HCC patients was analyzed using Cox proportional hazard analysis.The combination of 9 VOCs yielded 70.0% sensitivity, 88.6% specificity, and 75.0% accuracy for HCC diagnosis. When differentiating early HCC from cirrhotic patients, acetone dimer had a significantly higher AUC than AFP, i.e., 0.775 vs. 0.714, respectively, p = 0.001. Acetone dimer classified HCC patients into treatment responders and non-responders, with 95.7% sensitivity, 73.3% specificity, and 86.8% accuracy. Isopropyl alcohol was independently associated with the survival of HCC patients, with an adjusted hazard ratio of 7.23 (95%CI: 1.36-38.54), p = 0.020.Analysis of VOCs is a feasible noninvasive test for diagnosing and monitoring HCC treatment response.

    View details for DOI 10.3390/diagnostics13020257

    View details for Web of Science ID 000916965500001

    View details for PubMedID 36673067

    View details for PubMedCentralID PMC9858163

  • Circulating tumor cells as a prognostic biomarker in patients with hepatocellular carcinoma SCIENTIFIC REPORTS Prasoppokakorn, T., Buntho, A., Ingrungruanglert, P., Tiyarattanachai, T., Jaihan, T., Kulkraisri, K., Ariyaskul, D., Phathong, C., Israsena, N., Rerknimitr, R., Treeprasertsuk, S., Chaiteerakij, R. 2022; 12 (1): 18686

    Abstract

    Circulating tumor cells (CTCs) have been shown as a surrogate for cancer progression and prognostication. We aimed to determine an association between CTCs and survival of hepatocellular carcinoma (HCC) patients. Peripheral blood was obtained from 73 HCC patients to enumerate for epithelial CTCs/8 mL blood. CTCs were detected by immunoaffinity-based method using epithelial cell adhesion molecule (EpCAM) and mucin1 (MUC1). The CTCs detection rates of BCLC stages A, B, and C patients were 65.4% (17/26), 77.3% (17/22), and 96% (24/25), respectively, p = 0.018. Patients with CTCs < 5 cells/8 mL had significantly longer survival than those with CTCs ≥ 5 cells/8 mL (>36 vs. 4.6 months, p < 0.001). In multivariate analysis, CTP B, BCLC B, BCLC C, AFP ≥ 400 ng/mL, and CTC ≥ 5 cells/8 mL were independently associated with survival, with adjusted HRs (95%CI) of 4.1 (2.0-8.4), 3.5 (1.1-11.4), 4.7 (1.4-15.4), 2.4 (1.1-5.0), and 2.6 (1.2-8.4); p < 0.001, 0.036, 0.011, 0.025 and 0.012, respectively. The combination of CTCs ≥ 5 cells/8 mL and AFP ≥ 400 ng/mL provided additively increased HR to 5.3 (2.5-11.1), compared to HRs of 4.0 (2.0-8.0) and 3.5 (1.8-6.7) for CTCs ≥ 5 cells/8 mL and AFP ≥ 400 ng/mL, p < 0.001, respectively. The larger number of peripheral CTCs is correlated with higher tumor aggressive features and poorer survival of HCC patients. CTCs can potentially become novel prognostic biomarker in HCC.

    View details for DOI 10.1038/s41598-022-21888-9

    View details for Web of Science ID 000879109400037

    View details for PubMedID 36333384

    View details for PubMedCentralID PMC9636215

  • Positive predictive value of LI-RADS US-3 observations: multivariable analysis of clinical and imaging features. Abdominal radiology (New York) Tse, J. R., Shen, L., Tiyarattanachai, T., Bird, K. N., Liang, T., Yoon, L., Kamaya, A. 2022

    Abstract

    PURPOSE: To determine how clinical and imaging features affect the positive predictive values (PPV) of US-3 observations.METHODS: In this retrospective study, 10,546 adult patients who were high risk for hepatocellular carcinoma (HCC) from 2017 to 2021 underwent ultrasound screening/surveillance. Of these, 225 adult patients (100 women, 125 men) with an US-3 observation underwent diagnostic characterization with multiphasic CT (93; 41%), MRI (130; 58%), or contrast-enhanced ultrasound (2; 1%). US-3 observations included focal observations≥10mm in 216 patients and new venous thrombi in 9 patients. PPV with 95% confidence intervals were calculated using diagnostic characterization as the reference standard. Multivariable analysis of clinical and imaging features was performed to determine the strongest associations with cancer.RESULTS: Overall PPV for an US-3 observationwas 33% (27-39%) for at least intermediate probability of cancer (≥LR-3) and 15% (10-20%) for at least probable cancer (≥LR-4). At multivariable analysis, cirrhosis had the strongest effect size for at least probable cancer (p<0.001; odds ratio OR 20.4), followed by observation size (p<0.001; OR 2.65) and age (p=0.004; OR 1.05). Alpha-fetoprotein, visualization score, and observation echogenicity were not statistically significant associations. Modality (MRI versus CT) did not affect PPV. Due to the large effect of cirrhosis, PPV was then stratified by the presence (n=116; 52%) or absence (n=109; 48%) of cirrhosis. For at least probable cancer (≥LR-4), PPV increased from 4% (0-7%; non-cirrhotic) to 26% (18-34%; p<0.001; cirrhosis).CONCLUSION: Cirrhosis most strongly affects PPV of US-3 observations for at least probable cancer at diagnostic characterization among high-risk patients, increasing to 1 in 4 among cirrhotic patients from 1 in 25 among non-cirrhotic patients.

    View details for DOI 10.1007/s00261-022-03681-4

    View details for PubMedID 36253490

  • PERFORMANCE OF ABBREVIATED MAGNETIC RESONANCE IMAGING VERSUS ULTRASONOGRAPHY AS AN IMAGING TOOL FOR HEPATOCELLULAR CARCINOMA SURVEILLANCE Navadurong, H., Laohasurayotin, K., Yorwittaya, K., Tiyarattanachai, T., Tanpowpong, N., Pisuchpen, N., Chaiteerakij, R., Treeprasertsuk, S., Rerknimitr, R. BMJ PUBLISHING GROUP. 2022: A85
  • Interobserver agreement between eight observers using IOTA simple rules and O-RADS lexicon descriptors for adnexal masses. Abdominal radiology (New York) Antil, N., Raghu, P. R., Shen, L., Tiyarattanachai, T., Chang, E. M., Ferguson, C. W., Ho, A. A., Lutz, A. M., Mariano, A. J., Morimoto, L. N., Kamaya, A. 2022

    Abstract

    PURPOSE: To evaluate interobserver agreement in assigning imaging features and classifying adnexal masses using the IOTA simple rules versus O-RADS lexicon and identify causes of discrepancy.METHODS: Pelvic ultrasound (US) examinations in 114 women with 118 adnexal masses were evaluated by eight radiologists blinded to the final diagnosis (4 attendings and 4 fellows) using IOTA simple rules and O-RADS lexicon. Each feature category was analyzed for interobserver agreement using intraclass correlation coefficient (ICC) for ordinal variables and free marginal kappa for nominal variables. The two-tailed significance level (a) was set at 0.05.RESULTS: For IOTA simple rules, interobserver agreement was almost perfect for three malignant lesion categories (M2-4) and substantial for the remaining two (M1, M5) with k-values of 0.80-0.82 and 0.68-0.69, respectively. Interobserver agreement was almost perfect for two benign feature categories (B2, B3), substantial for two (B4, B5) and moderate for one (B1) with k-values of 0.81-0.90, 0.69-0.70 and 0.60, respectively. For O-RADS, interobserver agreement was almost perfect for two out of ten feature categories (ascites and peritoneal nodules) with k-values of 0.89 and 0.97. Interobserver agreement ranged from fair to substantial for the remaining eight feature categories with k-values of 0.39-0.61. Fellows and attendings had ICC values of 0.725 and 0.517, respectively.CONCLUSION: O-RADS had variable interobserver agreement with overall good agreement. IOTA simple rules had more uniform interobserver agreement with overall excellent agreement. Greater reader experience did not improve interobserver agreement with O-RADS.

    View details for DOI 10.1007/s00261-022-03580-8

    View details for PubMedID 35763052

  • Exhaled volatile organic compounds for diagnosis of hepatocellular carcinoma SCIENTIFIC REPORTS Sukaram, T., Tansawat, R., Apiparakoon, T., Tiyarattanachai, T., Marukatat, S., Rerknimitr, R., Chaiteerakij, R. 2022; 12 (1): 5326

    Abstract

    Volatile organic compounds (VOCs) profile for diagnosis and monitoring therapeutic response of hepatocellular carcinoma (HCC) has not been well studied. We determined VOCs profile in exhaled breath of 97 HCC patients and 111 controls using gas chromatography-mass spectrometry and Support Vector Machine algorithm. The combination of acetone, 1,4-pentadiene, methylene chloride, benzene, phenol and allyl methyl sulfide provided the highest accuracy of 79.6%, with 76.5% sensitivity and 82.7% specificity in the training set; and 55.4% accuracy, 44.0% sensitivity, and 75.0% specificity in the test set. This combination was correlated with the HCC stages demonstrating by the increased distance from the classification boundary when the stage advanced. For early HCC detection, d-limonene provided a 62.8% sensitivity, 51.8% specificity and 54.9% accuracy. The levels of acetone, butane and dimethyl sulfide were significantly altered after treatment. Patients with complete response had a greater decreased acetone level than those with remaining tumor post-treatment (73.38 ± 56.76 vs. 17.11 ± 58.86 (× 106 AU, p = 0.006). Using a cutoff of 35.9 × 106 AU, the reduction in acetone level predicted treatment response with 77.3% sensitivity, 83.3% specificity, 79.4%, accuracy, and AUC of 0.784. This study demonstrates the feasibility of exhaled VOCs as a non-invasive tool for diagnosis, monitoring of HCC progression and treatment response.

    View details for DOI 10.1038/s41598-022-08678-z

    View details for Web of Science ID 000775227200009

    View details for PubMedID 35351916

    View details for PubMedCentralID PMC8964758

  • Interpretable machine learning for characterization of focal liver lesions by contrast-enhanced ultrasound. IEEE transactions on ultrasonics, ferroelectrics, and frequency control Turco, S., Tiyarattanachai, T., Ebrahimkheil, K., Eisenbrey, J., Kamaya, A., Mischi, M., Lyshchik, A., El Kaffas, A. 2022; PP

    Abstract

    This work proposes an interpretable radiomics approach to differentiate between malignant and benign focal liver lesions (FLLs) on contrast-enhanced ultrasound (CEUS). Although CEUS has shown promise for differential FLLs diagnosis, current clinical assessment is performed only by qualitative analysis of the contrast enhancement patterns. Quantitative analysis is often hampered by the unavoidable presence of motion artefacts and by the complex, spatiotemporal nature of liver contrast enhancement, consisting of multiple, overlapping vascular phases. To fully exploit the wealth of information in CEUS, while coping with these challenges, here we propose to combine features extracted by temporal and spatiotemporal analysis in the arterial phase enhancement with spatial features extracted by texture analysis at different time points. Using the extracted features as input, several machine learning classifier are optimized to achieve semi-automatic FLLs characterization, for which there is no need for motion compensation and the only manual input required is the location of a suspicious lesion. Clinical validation on 87 FLLs from 72 patients at risk for HCC showed promising performance, achieving a balanced accuracy of 0.84 in the distinction between benign and malignant lesions. Analysis of feature relevance demonstrates that a combination of spatiotemporal and texture features is needed to achieve the best performance. Interpretation of the most relevant features suggests that aspects related to microvascular perfusion and the microvascular architecture, together with the spatial enhancement characteristics at wash-in and peak enhancement, are important to aid the accurate characterization of FLLs.

    View details for DOI 10.1109/TUFFC.2022.3161719

    View details for PubMedID 35320099

  • Application of artificial intelligence for diagnosis of pancreatic ductal adenocarcinoma by EUS: A systematic review and meta-analysis ENDOSCOPIC ULTRASOUND Prasoppokakorn, T., Tiyarattanachai, T., Chaiteerakij, R., Decharatanachart, P., Mekaroonkamol, P., Ridtitid, W., Kongkam, P., Rerknimitr, R. 2022; 11 (1): 17-26

    Abstract

    EUS-guided tissue acquisition carries certain risks from unnecessary needle puncture in the low-likelihood lesions. Artificial intelligence (AI) system may enable us to resolve these limitations. We aimed to assess the performance of AI-assisted diagnosis of pancreatic ductal adenocarcinoma (PDAC) by off-line evaluating the EUS images from different modes. The databases PubMed, EMBASE, SCOPUS, ISI, IEEE, and Association for Computing Machinery were systematically searched for relevant studies. The pooled sensitivity, specificity, diagnostic odds ratio (DOR), and summary receiver operating characteristic curve were estimated using R software. Of 369 publications, 8 studies with a total of 870 PDAC patients were included. The pooled sensitivity and specificity of AI-assisted EUS were 0.91 (95% confidence interval [CI], 0.87-0.93) and 0.90 (95% CI, 0.79-0.96), respectively, with DOR of 81.6 (95% CI, 32.2-207.3), for diagnosis of PDAC. The area under the curve was 0.923. AI-assisted B-mode EUS had pooled sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of 0.91, 0.90, 0.94, and 0.84, respectively; while AI-assisted contrast-enhanced EUS and AI-assisted EUS elastography had sensitivity, specificity, PPV, and NPV of 0.95, 0.95, 0.97, and 0.90; and 0.88, 0.83, 0.96 and 0.57, respectively. AI-assisted EUS has a high accuracy rate and may potentially enhance the performance of EUS by aiding the endosonographers to distinguish PDAC from other solid lesions. Validation of these findings in other independent cohorts and improvement of AI function as a real-time diagnosis to guide for tissue acquisition are warranted.

    View details for DOI 10.4103/EUS-D-20-00219

    View details for Web of Science ID 000761099300003

    View details for PubMedID 34937308

    View details for PubMedCentralID PMC8887033

  • Application of artificial intelligence in non-alcoholic fatty liver disease and liver fibrosis: a systematic review and meta-analysis THERAPEUTIC ADVANCES IN GASTROENTEROLOGY Decharatanachart, P., Chaiteerakij, R., Tiyarattanachai, T., Treeprasertsuk, S. 2021; 14: 17562848211062807

    Abstract

    The global prevalence of non-alcoholic fatty liver disease (NAFLD) continues to rise. Non-invasive diagnostic modalities including ultrasonography and clinical scoring systems have been proposed as alternatives to liver biopsy but with limited performance. Artificial intelligence (AI) is currently being integrated with conventional diagnostic methods in the hopes of performance improvements. We aimed to estimate the performance of AI-assisted systems for diagnosing NAFLD, non-alcoholic steatohepatitis (NASH), and liver fibrosis.A systematic review was performed to identify studies integrating AI in the diagnosis of NAFLD, NASH, and liver fibrosis. Pooled sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and summary receiver operating characteristic curves were calculated.Twenty-five studies were included in the systematic review. Meta-analysis of 13 studies showed that AI significantly improved the diagnosis of NAFLD, NASH and liver fibrosis. AI-assisted ultrasonography had excellent performance for diagnosing NAFLD, with a sensitivity, specificity, PPV, NPV of 0.97 (95% confidence interval (CI): 0.91-0.99), 0.98 (95% CI: 0.89-1.00), 0.98 (95% CI: 0.93-1.00), and 0.95 (95% CI: 0.88-0.98), respectively. The performance of AI-assisted ultrasonography was better than AI-assisted clinical data sets for the identification of NAFLD, which provided a sensitivity, specificity, PPV, NPV of 0.75 (95% CI: 0.66-0.82), 0.82 (95% CI: 0.74-0.88), 0.75 (95% CI: 0.60-0.86), and 0.82 (0.74-0.87), respectively. The area under the curves were 0.98 and 0.85 for AI-assisted ultrasonography and AI-assisted clinical data sets, respectively. AI-integrated clinical data sets had a pooled sensitivity, specificity of 0.80 (95%CI: 0.75-0.85), 0.69 (95%CI: 0.53-0.82) for identifying NASH, as well as 0.99-1.00 and 0.76-1.00 for diagnosing liver fibrosis stage F1-F4, respectively.AI-supported systems provide promising performance improvements for diagnosing NAFLD, NASH, and identifying liver fibrosis among NAFLD patients. Prospective trials with direct comparisons between AI-assisted modalities and conventional methods are warranted before real-world implementation.PROSPERO (CRD42021230391).

    View details for DOI 10.1177/17562848211062807

    View details for Web of Science ID 000734686700012

    View details for PubMedID 34987607

    View details for PubMedCentralID PMC8721422

  • Application of artificial intelligence in chronic liver diseases: a systematic review and meta-analysis BMC GASTROENTEROLOGY Decharatanachart, P., Chaiteerakij, R., Tiyarattanachai, T., Treeprasertsuk, S. 2021; 21 (1): 10

    Abstract

    The gold standard for the diagnosis of liver fibrosis and nonalcoholic fatty liver disease (NAFLD) is liver biopsy. Various noninvasive modalities, e.g., ultrasonography, elastography and clinical predictive scores, have been used as alternatives to liver biopsy, with limited performance. Recently, artificial intelligence (AI) models have been developed and integrated into noninvasive diagnostic tools to improve their performance.We systematically searched for studies on AI-assisted diagnosis of liver fibrosis and NAFLD on MEDLINE, Scopus, Web of Science and Google Scholar. The pooled sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and diagnostic odds ratio (DOR) with their 95% confidence intervals (95% CIs) were calculated using a random effects model. A summary receiver operating characteristic curve and the area under the curve was generated to determine the diagnostic accuracy of the AI-assisted system. Subgroup analyses by diagnostic modalities, population and AI classifiers were performed.We included 19 studies reporting the performances of AI-assisted ultrasonography, elastrography, computed tomography, magnetic resonance imaging and clinical parameters for the diagnosis of liver fibrosis and steatosis. For the diagnosis of liver fibrosis, the pooled sensitivity, specificity, PPV, NPV and DOR were 0.78 (0.71-0.85), 0.89 (0.81-0.94), 0.72 (0.58-0.83), 0.92 (0.88-0.94) and 31.58 (11.84-84.25), respectively, for cirrhosis; 0.86 (0.80-0.90), 0.87 (0.80-0.92), 0.85 (0.75-0.91), 0.88 (0.82-0.92) and 37.79 (16.01-89.19), respectively; for advanced fibrosis; and 0.86 (0.78-0.92), 0.81 (0.77-0.84), 0.88 (0.80-0.93), 0.77 (0.58-0.89) and 26.79 (14.47-49.62), respectively, for significant fibrosis. Subgroup analyses showed significant differences in performance for the diagnosis of fibrosis among different modalities. The pooled sensitivity, specificity, PPV, NPV and DOR were 0.97 (0.76-1.00), 0.91 (0.78-0.97), 0.95 (0.87-0.98), 0.93 (0.80-0.98) and 191.52 (38.82-944.81), respectively, for the diagnosis of liver steatosis.AI-assisted systems have promising potential for the diagnosis of liver fibrosis and NAFLD. Validations of their performances are warranted before implementing these AI-assisted systems in clinical practice.The protocol was registered with PROSPERO (CRD42020183295).

    View details for DOI 10.1186/s12876-020-01585-5

    View details for Web of Science ID 000608035600003

    View details for PubMedID 33407169

    View details for PubMedCentralID PMC7788739