Exploring learning transferability in deep segmentation of colorectal cancer liver metastases.

Development of Arterial Blood Supply in Experimental Liver Metastases

Revealing anti-PD-1 resistance mechanisms in HCC: A path towards novel combination immunotherapies

Stereotactic Body Radiation Therapy (SBRT) for Hepatoma and Liver Metastasis: Evaluation of the Optimal Dose

Society of Interventional Radiology Position Statement on Chemoembolization of Hepatic Malignancies

Thymidine phosphorylase (TP) and dihydropyrimidine dehydrogenase (DPD) are considered to be key enzymes affecting the prognosis for patients with various cancers. We tried to prove the correlation of TP and DPD expression in hepatocellular carcinoma (HCC) and liver metastasis. We quantified TP and DPD levels by an enzyme-linked immunosorbent assay (ELISA) in the tumor (T) and adjacent normal tissue (N) obtained from 8 HCC patients, and 11 liver metastasis patients together with 9 of their primary cancers. TP levels were higher in the primary cancer, liver metastasis, and HCC compared with each adjacent tissue. TP levels were higher in HCC than in liver metastasis, and TP levels in the adjacent tissues of HCC were also higher than those in adjacent tissues of liver metastasis. TP levels were higher in liver metastasis than in primary cancer, and TP levels in adjacent tissues of liver metastasis were also higher than those in adjacent tissues of primary cancer. However, there were no differences in TP T/N ratio between HCC and liver metastasis, and between primary cancer and liver metastasis. DPD levels were lower in the liver metastasis compared with the adjacent liver tissues, and DPD levels in liver metastasis or its adjacent liver tissues were higher than those in primary cancer or its adjacent tissues. There were no differences in DPD T/N ratio between HCC and liver metastasis, and between primary cancer and liver metastasis. Thus, we demonstrated that TP was highly expressed in liver malignancy. We may be able to increase the success of anticancer chemotherapy for liver malignancy while decreasing the side effects by analysis of T/N ratios in TP, DPD, and TP/DPD in addition to TP expression.

New frontiers in liver resection for hepatocellular carcinoma.

Nowadays, deep features, obtained from a variety of deep learning architectures, play an important role in several real problems. It is know that transfer learning strategies could be employed to take advantage of such deep features trained under a general context (e.g. ImageNet). However, to the best of our knowledge, the majority of works focus on similar contexts to accomplish such transfer strategies. Thus, in this work we analyze the role of deep features in content-based medical image retrieval, and demonstrate that it is possible to make use of transfer learning from a general context to a specific medical context, like the content-based mammographic image retrieval. To do so, we evaluated several hand-crafted features against deep features acquired from state-of-the-art deep architectures through transfer learning. Extensive experiments on challenging public mammographic image datasets testify that the generalized deep features are able to improve in a great extend the precision of similarity queries both in the traditional process and applying query refinement strategies.

Locoregional relapse occurs in over half of gastric cancer patients who undergo potentially curative resection. Adjuvant chemoradiation reduces locoregional relapse, but often requires irradiating large fields and is limited by poor patient tolerance. This study explores the potential dosimetric benefit in reducing the radiation dose to normal structures by treating gastroesophageal (GE) junction/proximal gastric cancers with preoperative rather than adjuvant radiotherapy. Five cases of GE junction/proximal gastric cancer patients treated postoperatively with curative intent were selected. The actual target contours were then modified to reflect hypothetical target volumes which would have been used had the patients been treated preoperatively. Hypothetical preoperative treatment plans were generated for each patient based on these modified contours. The hypothetical preoperative treatment plans were then compared to the actual postoperative plans with respect to dose-volume parameters including lung mean dose, lung V20, heart V20 and V30, and mean doses to abdominal structures. Target volumes were smaller with preoperative treatment, with an average reduction of 23%. Comparative dose-volume histogram (DVH) analysis showed the resultant composite lung doses were reduced in the preoperative plans by 50-79%. In all patients, the proportion of lungs receiving at least 20 Gy (V20) was substantially reduced using preoperative treatment (1.9% vs. 9.7% in the 3-D conformal patient; mean of 3.1% vs. 17.6% in the intensity modulated radiation therapy patients). Likewise, the volume of heart receiving at least 30 Gy was dramatically reduced in all preoperative plans (15.8% vs. 35.4%). Doses to the kidneys, liver and spinal cord were comparable in both approaches. Preoperative treatment of GE junction and proximal gastric cancer patients offers the potential to decrease the radiation dose received by normal thoracic structures.

489P Analysis of 3-dimensional volumetric distribution and dispersion of the radioenhancer NBTXR3 in various solid malignancies

4590 Background: Safety of stereotactic radiotherapy (SRT) for unresectable hepatocellular carcinoma (HCC), intrahepatic cholangiocarcinoma (CC) and liver metastases (LM) has not been well established. Results from a phase I study of 6 fraction SRT are reported. Methods: Eligible patients had unresectable or medically inoperable HCC, CC or LM, liver enzymes < 6 fold higher than normal, platelets > 80,000 bil/L, Child score A, > 800 cc uninvolved liver, KPS = 60 and unsuitability for standard therapies. Patients were treated with breath hold and image guided 6 fraction SRT. Dose was individualized to maintain the same risk of liver toxicity at three risk levels (I-5%, II-10%, III-20%). Escalation to level II and III occurred once at least 3 patients had been followed for >3 months without dose limiting toxicity (grade 4/5 < 1 month or grade 4/5 liver < 3 months) for each stratum. Stratification was based on diagnosis and liver volume irradiated (low <20%, mid 20–50%, high 50–80%). Results: From Aug. 2003 to Dec. 2006, 82 patients initiated SRT. Two patients discontinued SRT after 1 fraction for progressive disease (LM) and a variceal bleed (HCC). 80 patients completed SRT (38 LM, 32 HCC, 10 CC). Median age was 64 years (38–92 years). Median tumor volume was 293 cc (3–3088 cc). 24 patients (30%) had extra-hepatic disease. 17 HCC patients had portal vein thrombosis (53%), 14 HBV, 12 HCV and 4 alcoholic cirrhosis. The median prescribed dose was 40 Gy (24 Gy - 60 Gy) in 6 fractions. Within 3 months post SRT, no dose-limiting grade 4/5 toxicity or classic radiation liver toxicity was observed. Grade 3 liver enzymes (2 new, 9 pre-existing), thrombocytopenia (3), nausea (3) and fatigue (1) was observed. Child score declined in 8 patients (5 HCC, 2 CC, 1 LM), 6 with progressive disease. Late toxicity included 1 tumor-duodenal fistula, 1 bowel obstruction and 2 GI bleeds. The in-field response rate was 60% (LM 57%, CC 50%, HCC 67%): CR 14%, PR 46%, SD 23%, PD 17%. Actuarial 12 month local control was 78% (95% CI: 58- 90%). The median survival for LM, CC and HCC was 16.6 months (7.9–25.6), 13.1 months (6.0–28.4) and 11.0 months (8.6–20.8) respectively. Conclusions: Individualized SRT is a safe, promising treatment for unresectable liver cancer. No significant financial relationships to disclose.

To investigate the diagnostic performance of acoustic radiation force impulse (ARFI) elastography for characterizing focal liver mass by quantifying their stiffness. This prospective study included 62 patients with a focal liver mass that was well visualized on conventional ultrasonography performed in our institution from February 2011 to November 2011. Among them, 12 patients were excluded for ARFI measurement failure due to a lesion that was smaller than the region of the interest and at an inaccessible location (deeper than 8 cm) (n = 7) or poor compliance to hold their breath as required (n = 5). Finally, 50 patients with valid ARFI measurements were enrolled. If a patient had multiple liver masses, only one mass of interest was chosen. The masses were diagnosed by histological examination or clinical diagnostic criteria. During ultrasonographic evaluation, stiffness, expressed as velocity, was checked 10 times per focal liver mass and the surrounding liver parenchyma. After further excluding three masses that were non-diagnostic on biopsy, a total of 47 focal mass lesions were tested, including 39 (83.0%) malignant masses [24 hepatocellular carcinomas (HCC), seven cholangiocellular carcinomas (CCC), and eight liver metastases] and eight (17.0%) benign masses (five hemangiomas and three focal nodular hyperplasias, FNH). Thirty-seven (74.0%) masses were confirmed by histological examination. The mean velocity was 2.48 m/s in HCCs, 1.65 m/s in CCCs, 2.35 m/s in metastases, 1.83 m/s in hemangiomas, and 0.97 m/s in FNHs. Although considerable overlap was still noted between malignant and benign masses, significant differences in ARFI values were observed between malignant and benign masses (mean 2.31 m/s vs 1.51 m/s, P = 0.047), as well as between HCCs and benign masses (mean 2.48 m/s vs 1.51 m/s, P = 0.006). The areas under the receiver operating characteristics curves (AUROC) for discriminating the malignant masses from benign masses was 0.724 (95%CI, 0.566-0.883, P = 0.048), and the AUROC for discriminating HCCs from benign masses was 0.813 (95%CI, 0.649-0.976, P = 0.008). To maximize the sum of sensitivity and specificity, an ARFI value of 1.82 m/s was selected as the cutoff value to differentiate malignant from benign liver masses. Furthermore, the cutoff value for distinguishing HCCs from benign masses was also determined to be 1.82 m/s. The diagnostic performance of the sum of the ARFI values for focal liver masses and the surrounding liver parenchyma to differentiate liver masses improved (AUROC = 0.853; 95%CI, 0.745-0.960; P = 0.002 in malignant liver masses vs benign ones and AUROC = 0.948; 95%CI, 0.896-0.992, P < 0.001 in HCCs vs benign masses). ARFI elastography provides additional information for the differential diagnosis of liver masses. However, our results should be interpreted in clinical context, because considerable overlap in ARFI values existed among liver masses.

Hepatocellular carcinoma (HCC) is the seventh most common cancer worldwide and the third most common cause of cancer-related deaths; the number of new cases per year is approaching 750,000. The magnitude of the incidence of HCC has discouraged any attempts to apply liver transplantation (LT) as the prevailing curative therapy for HCC worldwide because of the limited sources of donated organs (deceased and living donors) and the poor access to sophisticated health care systems in some geographical areas. If these limitations continue to prevail throughout the world, any attempt to significantly reduce HCC-related mortality rates through the application of LT will be delusional. International experiences have confirmed, however, the potential of LT to definitively cure HCC because it presents a unique opportunity to remove both the tumor (HCC is associated with 695,000 deaths per year) and the underlying cirrhosis. Despite its limited access, LT has become the standard of care for patients with small HCCs and the main driving force for alternative strategies offered to patients with intermediate HCCs. In 1996, a prospective cohort study defined restrictive selection criteria that led to superior survival for transplant patients in comparison with any other previous experience with transplantation or other options for HCC. Since then, these selection criteria have become universally known as the Milan criteria (MC) in recognition of their origin. Ever since their adoption in clinical practice, the MC have helped doctors to single out early-stage HCC as a prognostic category of cancer presentation that is amenable to curative treatments. After their implementation, the favorable posttransplant outcomes that were observed in cohort series were so convincing that the MC immediately became the standard of care for early HCC, and further validation by randomized controlled trials (RCTs) was prevented. After the passage of approximately a decade, researchers began to challenge the MC with other proposals designed to capture those patients not meeting the MC who could achieve similar posttransplant survival rates through the expansion of the accepted tumor limits for transplant eligibility. None of these expanded criteria have become the new reference standard for selecting LT candidates with HCC; any broadening of the selection criteria for transplantation is inevitably hampered by severe

Automatic medical image segmentation has become increasingly important as contemporary medical imaging has become more widely available and used. Existing image segmentation solutions however lack the necessary functionality for simple medical image segmentation pipeline design. Pipelines that have already been deployed are frequently standalone software that has been optimised for a certain public data collection. As a result, the open-source python module deep-Convolutional neural network-Restricted Boltzmann Machine (deep CNNRBM) was introduced in this research work. The goal of Deep CNN-purpose RBMs is to have an easy-touse API that allows for the rapid creation of medical image segmentation transmission lines that include data augmentation, metrics, data I/O pre-processing, patch wise analysis, a library of pre-built deep neural networks, and fully automated assessment. Similarly, comprehensive pipeline customisation is possible because of strong configurability and many open interfaces. The dataset of Kidney tumor Segmentation challenge 2019 (KiTS19) acquired a strong predictor with respect to the standard 3D U-net model after cross-validation using deep CNNRBM. To that purpose, deep CNN-RBM, an expressive deep learning medical image segmentation architecture is introduced. The CNN sub-model captures frame-level spatial features automatically while the RBM submodel fuses spatial data over time to learn higher-level semantics in kidney tumor prediction. A neural network recognises medical picture segmentation, which is initiated using RBM to second-order collected data and then fine-tuned using back propagation to be more differential. According to the simulation outcome, the proposed deep CNN-RBM produced good classification results on the kidney tumour segmentation dataset.

Purpose: To assess whether ring enhancements of liver metastases on Mn-DPDP enhanced, early MR images were well visualized on delayed images, as compared with those of hepatocellular carcinomas (HCC), and to investigate the detection accuracy and conspicuity of each tumor. Materials and Methods: Twenty patients with liver metastases and 15 with HCC were studied by Mn-DPDP enhanced, T1-weighted MR images. Peripheral ring enhancement and conspicuity were investigated. Differences in detection accuracy and frequency of ring enhancement in liver metastases and HCC were assessed. Results: In liver metastases (n=69), 44 cases (63.8%) without ring enhancement and 25 (36.2%) with were noted on early images. Sixteen cases (23.2%) without ring enhancement, 38 (55.1%) with ring enhancement similar to the early images, and 15 (21.7%) with prominent ring enhancement were noted on delayed images. In HCC (n=37), 36 cases (97.3%) without ring enhancement and 1 case (2.7%) with were noted on early images. There was no difference of detection accuracy in liver metastases or HCC between the 2 readings. Ring enhancement and conspicuity of each tumor were superior on delayed images. Ring enhancement in liver metastases was better seen on delayed images. Conclusion: Ring enhancement in liver metastases was well presented on Mn-DPDP enhanced, delayed MR images, which was useful to differentiate liver metastases from HCC.

PURPOSE: The indication for liver transplantation in malignant liver tumors has been controversial due to disappointing results and shortage of donor organs. The authors evaluated the experience and results of a single center in order to define present indications and selection criteria in hepatobiliary malignancy. PATIENTS AND METHODS: Retrospective analysis of 212 patients who underwent liver transplantation for malignant tumors between 1972 and 1995: Primary hepatobiliary tumors: hepatocellular carcinoma, n = 124 (with underlying cirrhosis, n = 86; fibrolamellar subtype, n = 8); intrahepatic bile duct (cholangiocellular) carcinoma, n = 24; proximal bile duct carcinoma, n = 29; other uncommon entities (n = 15); secondary liver tumors: neuroendocrine, n = 11, and nonendocrine, n = 9. RESULTS: Survival rates in primary liver cancer were correlated to International Union Against Cancer (UICC) tumor stage. For hepatocellular and proximal bile duct carcinoma significantly better outcome was found in UICC-tumor stage I and II versus III and IV. No long-term survival was found after transplantation for intrahepatic bile duct carcinoma, hemangiosarcoma and nonendocrine liver metastases. Comparison of transplant and resected patients with hepatocellular carcinoma stage I and II with underlying cirrhosis showed better survival after transplantation: 1-, 3-, 5-year survival rate of 83.3% versus 76.9%, 75.8% versus 44.0%, 60.6% versus 44.0%, and median survival 96.5 versus 23.2 months. Although this difference was not significant, no patient died from tumor recurrence in the transplant group versus three in the resection group. DISCUSSION AND CONCLUSIONS: Patients with malignant tumors can be selected for transplantation with predictable likelihood for long-term survival. According to the present data, liver transplantation can be considered in unresectable UICC-stage II hepatocellular and proximal bile duct carcinoma, the uncommon entities fibrolamellar carcinoma, epitheliod hemangioendothelioma and hepatoblastoma as well as liver metastases from neuroendocrine tumors. UICC-stage II and IV hepatocellular carcinoma as well as intrahepatic bile duct carcinoma, hemangiosarcoma and metastases from nonendocrine tumors should be excluded from transplantation alone. For hepatocellular carcinoma, multimodality treatment protocols have had a proven impact on the prevention of early recurrence and prolongation of survival. There is evidence that liver transplantation in still resectable hepatocellular carcinoma with underlying cirrhosis might be more appropriate in order to cure the cancer-bearing disease.