ROMA Algorithm Research Articles

An ML-based decision support system for reliable diagnosis of ovarian cancer by leveraging explainable AI

Ovarian cancer (OC) is one of the most prevalent types of cancer in women. Early and accurate diagnosis is crucial for the survival of the patients. However, the majority of women are diagnosed in advanced stages due to the lack of effective biomarkers and accurate screening tools. While previous studies sought a common biomarker, our study suggests different biomarkers for the premenopausal and postmenopausal populations. This can provide a new perspective in the search for novel predictors for the effective diagnosis of OC. Genetic algorithm has been utilized to identify the most significant biomarkers. The XGBoost classifier is then trained on the selected features and high ROC-AUC scores of 0.864 and 0.911 have been obtained for the premenopausal and postmenopausal populations, respectively. Lack of explainability is one major limitation of current AI systems. The stochastic nature of the ML algorithms raises concerns about the reliability of the system as it is difficult to interpret the reasons behind the decisions. To increase the trustworthiness and accountability of the diagnostic system as well as to provide transparency and explanations behind the predictions, explainable AI has been incorporated into the ML framework. SHAP is employed to quantify the contributions of the selected biomarkers and determine the most discriminative features. Merging SHAP with the ML models enables clinicians to investigate individual decisions made by the model and gain insights into the factors leading to that prediction. Thus, a hybrid decision support system has been established that can eliminate the bottlenecks caused by the black-box nature of the ML algorithms providing a safe and trustworthy AI tool. The diagnostic accuracy obtained from the proposed system outperforms the existing methods as well as the state-of-the-art ROMA algorithm by a substantial margin which signifies its potential to be an effective tool in the differential diagnosis of OC.

Diagnostic Utility of Selected Matrix Metalloproteinases (MMP-2, MMP-3, MMP-11, MMP-26), HE4, CA125 and ROMA Algorithm in Diagnosis of Ovarian Cancer.

Ovarian cancer (OC) has an unfavorable prognosis. Due to the lack of effective screening tests, new diagnostic methods are being sought to detect OC earlier. The aim of this study was to evaluate the concentration and diagnostic utility of selected matrix metalloproteinases (MMPs) as OC markers in comparison with HE4, CA125 and the ROMA algorithm. The study group consisted of 120 patients with OC; the comparison group consisted of 70 patients with benign lesions and 50 healthy women. MMPs were determined via the ELISA method, HE4 and CA125 by CMIA. Patients with OC had elevated levels of MMP-3 and MMP-11, similar to HE4, CA125 and ROMA values. The highest SE, SP, NPV and PPV values were found for MMP-26, CA125 and ROMA in OC patients. Performing combined analyses of ROMA with selected MMPs increased the values of diagnostic parameters. The topmost diagnostic power of the test was obtained for MMP-26, CA125, HE4 and ROMA and performing combined analyses of MMPs and ROMA enhanced the diagnostic power of the test. The obtained results indicate that the tested MMPs do not show potential as stand-alone OC biomarkers, but can be considered as additional tests to raise the diagnostic utility of the ROMA algorithm.

Krukenberg tumour in a patient with secondary infertility – case report

A patient with secondary infertility, diagnosed with an endometrial polyp and right ovarian cyst, was referred for hysterolaparoscopy treatment. The woman reported no symptoms. During the hospitalisation, in addition to the aforementioned issues, left ovarian lesions and ascites were identified. A 10 cm lesion in the right ovary exhibited abundant vascularity. Tumour markers (AFP, CEA, HE4, CA-125) and the ROMA algorithm were assessed and found to be within the normal range. Laparotomy revealed a high-grade malignant tumour of unknown origin confirmed by histological examination. Uterine resection with unchanged greater omentum was performed. Immunohistochemical tests revealed positive cytokeratin (CK) reaction, including CK7 and focal CK20, as well as a marker of proliferation Ki-67 in some cells (20–30%). Signet ring cells and positive mucicarmine stain reaction were detected. The histological evaluation confirmed a Krukenberg tumour originating most likely from the stomach. The patient was referred to the Maria Skłodowska-Curie Greater Poland Cancer for further treatment.

Performance of IOTA Simple Rules Risks, ADNEX Model, Subjective Assessment Compared to CA125 and HE4 with ROMA Algorithm in Discriminating between Benign, Borderline and Stage I Malignant Adnexal Lesions.

Borderline ovarian tumors (BOTs) and early clinical stage malignant adnexal masses can make sonographic diagnosis challenging, while the clinical utility of tumor markers, e.g., CA125 and HE4, or the ROMA algorithm, remains controversial in such cases. To compare the IOTA group Simple Rules Risk (SRR), the ADNEX model and the subjective assessment (SA) with serum CA125, HE4 and the ROMA algorithm in the preoperative discrimination between benign tumors, BOTs and stage I malignant ovarian lesions (MOLs). A multicenter retrospective study was conducted with lesions classified prospectively using subjective assessment and tumor markers with the ROMA. The SRR assessment and ADNEX risk estimation were applied retrospectively. The sensitivity, specificity, positive and negative likelihood ratios (LR+ and LR-) were calculated for all tests. In total, 108 patients (the median age: 48 yrs, 44 postmenopausal) with 62 (79.6%) benign masses, 26 (24.1%) BOTs and 20 (18.5%) stage I MOLs were included. When comparing benign masses with combined BOTs and stage I MOLs, SA correctly identified 76% of benign masses, 69% of BOTs and 80% of stage I MOLs. Significant differences were found for the presence and size of the largest solid component (p = 0.0006), the number of papillary projections (p = 0.01), papillation contour (p = 0.008) and IOTA color score (p = 0.0009). The SRR and ADNEX models were characterized by the highest sensitivity (80% and 70%, respectively), whereas the highest specificity was found for SA (94%). The corresponding likelihood ratios were as follows: LR+ = 3.59 and LR- = 0.43 for the ADNEX; LR+ = 6.40 and LR- = 0.63 for SA and LR+ = 1.85 with LR- = 0.35 for the SRR. The sensitivity and specificity of the ROMA test were 50% and 85%, respectively, with LR+ = 3.44 and LR- = 0.58. Of all the tests, the ADNEX model had the highest diagnostic accuracy of 76%. This study demonstrates the limited value of diagnostics based on CA125 and HE4 serum tumor markers and the ROMA algorithm as independent modalities for the detection of BOTs and early stage adnexal malignant tumors in women. SA and IOTA methods based on ultrasound examination may present superior value over tumor marker assessment.

Peritoneal inclusion cysts as a diagnostic and treatment challenge.

Peritoneal inclusion cysts (PICs) are benign multilocular cysts that consist of fluid localized between intraperitoneal adhesions. They usually present in women in the third and fourth decades of life with a history of prior pelvic or abdominal surgery, especially months to 20 years ago. PICs have low mortality and the potential for high morbidity. Transvaginal ultrasound with color Doppler has an important role in the differential diagnosis. Also, a very important role is played by CA 125 plasma level or use of the ROMA algorithm. But thanks to laparoscopy and the possibility of biopsy from suspicious lesions the correct diagnosis can be established. Treatment of PICs depends individually on the patient's condition, symptoms, other diseases, and desire for procreation. Among other contemplated are hormonal oral contraceptive, an image-guided aspiration, minimally invasive or open surgery. The issue of diagnosis and treatment of PICs requires the continuation of multicentre, randomized clinical trials to find and standardize effective, personalized treatments for PICs.

Is HE4 Superior over CA-125 in the Follow-up of Patients with Epithelial Ovarian Cancer?

Notwithstanding important advances in the treatment of epithelial ovarian cancer (EOC), this disease is still a leading cause of global high mortality from gynecological malignancies. Recurrence in EOC is inevitable and it is responsible for poor survival rates. There is a critical need for novel effective biomarkers with improved accuracy compared to the standard carbohydrate antigen-125 (CA-125) for follow-up. The human epididymis protein 4 (HE4) is used for early detection of EOC (ROMA algorithm) as well as for predicting optimal cytoreduction after neoadjuvant chemotherapy and survival outcomes. Notably, the emerging HE4 is a promising prognostic biomarker that has displayed better accuracy in various recent studies for detecting recurrent disease. In this mini-review, we discussed the potential of HE4 as an accurate predictor of EOC recurrence.

HE4 – not only an ovarian cancer biomarker – a brief review

Human epididymis protein 4 (HE4) was firstly identified in epididymal epithelial cells and described as a protease inhibitor playing a role in spermatogenesis. Regarding numerous studies proving its diverse potential as a prognostic and predictive factor in ovarian cancer, it was incorporated into ROMA algorithm. Nevertheless, recent studies have shown that serum level of HE4 is not exclusive to ovarian cancer. As a result, doctors using ROMA algorithm for stratifying patients with ovarian cancer must be aware of other conditions that may affect serum level of HE4. This review comprises different conditions connected with high level of HE4 that might impact ovarian cancer diagnosing process. Moreover, discovering increased HE4 level in various conditions should open discussion about its applicability in diseases other than ovarian cancer.

Optimal cosmic microwave background map-making in the presence of cross-correlated noise

We present an extension of the ROMA map-making algorithm for the generation of optimal cosmic microwave background polarization maps. The new code allows for a possible cross-correlated noise component among the detectors of a CMB experiment. A promising application is the forthcoming LSPE balloon-borne experiment, which is devoted to the accurate observation of CMB polarization at large angular scales. We generalized the noise covariance matrix in time domain to account for all the off-diagonal terms due to the detector cross-talk. Hence, we performed preliminary forecasts of the LSPE-SWIPE instrument. We found that considering the noise cross-correlation among the detectors results in a more realistic estimate of the angular power spectra. In particular, the extended ROMA algorithm has provided a considerable reduction of the spectra error bars. We expect that this improvement could be crucial in constraining the B-mode polarization at the largest scales.

Preoperative HE4, CA125 and ROMA in the differential diagnosis of benign and malignant adnexal masses.

BackgroundThe aim of this study was to evaluate HE4, CA125 and ROMA in the preoperative differentiation benign ovarian diseases from epithelial ovarian cancer depending on the menopausal status.MethodsIn order to estimate markers’ concentrations in the serum of women with benign ovarian disease (n = 128) and with epithelial ovarian carcinoma (n = 96) the electrochemiluminescence (ECLIA) technique has been applied.ResultsUsing the ROC analysis, although no statistical differences were found among their AUCs, the ROMA algorithm seems to be effective in gathering the diverse performance of HE4 and CA125. The AUC for HE4, CA125 and ROMA for all patients were: 0.895; 0.879 and 0.918, respectively. At established new optimal cutoff values for HE4, CA125 and ROMA we found higher specificity in postmenopausal compared to premenopausal women (96.9 vs 89.8 % and 97.7 vs 84.1 % and 95.9 vs 89.1 %, respectively). The sensitivity of HE4 in pre- and postmenopausal women was similar (83.5 vs 83.8 %), while for CA125 was the highest in premenopausal women (87.0 vs 84.1 %). For HE4, CA125 and ROMA the negative predictive value was high (97.6, 93.9 and 94.4 %, respectively).ConclusionsThe ROMA algorithm shows the best diagnostic performance to distinguish epithelial ovarian cancer from benign ovarian disease. We found the high specificity of HE4 and CA125 while differentiating ovarian benign diseases from epithelial ovarian cancer in postmenopausal women and the high sensitivity of CA125 in detecting epithelial ovarian cancer in premenopausal patients.

Evaluation of applicability of HE4 and ROMA in the preoperative diagnosis of adnexal masses

The aim of the study was to evaluate the effectiveness of HE4 alone and in combination with CA 125 (ROMA) in selecting patients at high risk of adnexal malignancy. Serum CA 125 and HE4 levels were determined and the ROMA value was calculated in 259 women qualified for surgery due to adnexal mass. The results were compared with histopathological findings. Sensitivity and specificity in preoperative diagnosis of primary ovarian cancer were 93.2% and 71.5% for CA 125 and 95.4% and 81.3% for HE4, respectively ROMA algorithm achieved sensitivity of 95.4% and specificity of 79.8%. All methods reached sensitivity of 100% at specificity of 65.6% for CA125, 93.4% for HE4 and 82.0% for ROMA in premenopausal women, whereas in postmenopausal women sensitivity and specificity achieved levels of 92.1% and 81.7% for CA 125, 94.7% and 60.6% for HE4 and 94.7% and 76.1% for ROMA, respectively Serum levels of both CA 125 and HE4 were significantly higher in women with primary ovarian cancer as compared to benign disease. Concentrations of CA 125 in patients with endometriosis were significantly elevated as compared to women with other benign tumors. Such relation was not observed when HE4 levels were concerned. CA 125, HE4 and ROMA are useful in preoperative diagnosis of ovarian malignancy HE4 improves the diagnostic accuracy in cases of endometriosis, verifying false positive results of CA 125.

An analysis of preoperatively HE4 and CA125 in matched samples of patients (pts) with benign disease and borderline ovarian tumors (BOT) with and without invasive implants (InvImp).

e22102 Background: Presence of InvImp, quality of surgical staging and residual tumor mass were shown to be major prognostic factors for BOT. Currently, there are no biomarkers that can predict the various types of BOT. Aim of our study was to assess the value of individual CA125 and HE4 values, and the combination of CA125 and HE4 according to the ROMA algorithm for detecting BOT, and for predicting the presence of InvImp. Methods: BOT pts with InvImp (n=14) were matched by age and histology and compared to consecutive pts with benign disease and BOT without InvImp (1 to 4 to 1, respectively). Blood was collected prior surgery. HE4 and CA125 were determined using ELISA. The performance of HE4 and CA125 were analyzed using the receiver operator characteristic (ROC) and logistic regression. Results: HE4 and CA125 were increased in BOT with InvImp (median=73.85pM and 49.5IU/dl) compared with benign diseases and BOT without InvImp (median values: 48pM and 18.5IU/dl, and 61.45pM and 26.5IU/dl, respectively). HE4 and CA125 expression were significantly increased in BOT with InvImp compared to benign diseases (p=0.029 and 0.004, respectively) but not to BOT without InvImp. Furthermore we grouped BOT without InvImp together with benign diseases and compared them with BOT with InvImp. HE4 and CA125 cut off values of 65pM and 60IU/dl, respectively were independent predictive factors for the presence of InvImp (p=0.018, OR=5.32, 95%CI 1.33-21.20 and p=0.031, OR=4.83, 95%CI=1.16-20.13, respectively). ROMA performed better than HE4 or CA125 alone (p=0.002, RR=9.17, 95%CI=2.31-36.36). The OR of a high ROMA score were 0.082 (95%CI: 0.02-0.34; p=0.001) for benign diseases and 0.27 (95%CI: 0.05-1.42; p=0.123) for BOT without InvImp compared to BOT with InvImp. Conclusions: The ROMA score can not predict sufficiently InvImp in BOT. Further multicenter and larger prospective studies are needed.

Comparison of HE4, CA125 and ROMA algorithm in women with a pelvic mass: Correlation with pathological outcome

ObjectiveThe quality of first surgery is one of the most important prognostic factors in ovarian cancer patients. Pre-surgical distinction of benign and malignant pelvic mass plays a critical role in ovarian cancer management and survival. The aim of this study was to evaluate the clinical performance of ROMA algorithm and of CA125 and HE4 in the triage of patients with a pelvic mass undergoing surgery, in order to discriminate benign from malignant disease. MethodsThree hundred and forty-nine pre- and post-menopausal women, aged 18years or older undergoing surgery because of a pelvic mass were enrolled: serum concentrations of CA125 and HE4 were determined and ROMA was calculated for each sample. ResultsMedian serum CA125 and HE4 levels were higher in patients with EOC compared to subjects with benign disease (p<0.0001). The resultant accuracy (using Receiver Operating Characteristics, ROC Area) values for HE4, CA125 and ROMA showed a good performance ranging from 89.8% for CA125 in pre-menopausal patients to 93.3% for ROMA in post-menopausal patients: AUC for ROMA resulted significantly higher in comparison to CA125 alone (93.3% vs 90.3%, p=0.0018) in post menopausal patients. A sub-analysis considering the 40 patients with endometrioid disease showed the highest accuracy of HE4 in these patients. ConclusionsData presented confirm the accuracy of HE4 and of the ROMA algorithm in the distinction of ovarian carcinoma from benign disease, with a trend towards better performance for ROMA than for CA125 alone, statistically significant in postmenopausal patients.

High preoperative blood levels of HE4 predicts poor prognosis in patients with ovarian cancer

The aim of this study was to assess the clinical value of preoperative blood levels of HE4 as a predictor of overall survival in patients with ovarian cancer and to validate previous data of HE4 and the ROMA algorithm including HE4 and CA125 in discriminating benign and malignant ovarian tumors.Experimental designThe preoperative plasma levels of HE4 and CA125 were analyzed with ELISA in 312 patients with adnexal lesions. Tumors were classified as benign (n= 206), borderline (i.e. low malignant potential tumors) (n= 25), and well (n= 14), moderately (n= 15), and poorly (n= 51) differentiated malignant.ResultsIn univariate Cox regression analyses high levels (dichotomized at the median) of HE4, CA125, increased age (continuous variable), advanced-stage of disease 2–4, histological grade 3 and non-optimal tumor debulking at primary surgery were all significantly associated with shorter overall survival. A multivariate Cox regression model including pre-operative available covariates HE4 and CA125 both dichotomized at median in addition to age as continuous variable showed that high levels of HE4 was an independent prognostic marker for worse prognosis HR 2.02 (95% CI 1.1-3.8). In postmenopausal women the ROMA algorithm gave the highest AUC of 0.94 (95% CI, 0.90-0.97) which was higher than the separate markers HE4 AUC 0.91 (95% CI 0.86-0.95) and CA125 AUC 0.91(95% CI 0.87-0.96).ConclusionsHigh concentration of plasma HE4 is an independent preoperative marker of poor prognosis in patients with ovarian cancer. The algorithm ROMA discriminates in postmenopausal women between malignant and benign tumors with an AUC of 0.94.

HE4, CA-125, and cystic ovarian mass

To the editor: The recent publication on 'HE4, CA-125, and cystic ovarian mass' is very interesting [1]. Partheen et al. [1] concluded that HE4 did not outperform CA-125. Indeed, the use of two biomarkers might be expected for increased screening ability; however, the problem of the cost should be kept in mind. Incremental analysis on additional cost should be done. Based on the present report, the question is whether it should recommend screening for both HE4 and CA-125 in the patient presenting with a suspicious mass at ovary. This topic is the further point for future studies as indicated in many recent studies [1,2]. Finally, the use of additional means, such as ROMA algorithm and symptom index, to add more value of using HE4 and CA-125 screening should also be further assessed [2,3].

Reply: The performance of the risk of ovarian malignancy algorithm

Reply: The performance of the risk of ovarian malignancy algorithm

HE4 and epithelial ovarian cancer: Comparison and clinical evaluation of two immunoassays and a combination algorithm

BackgroundTwo commercial immunoassays for HE4 have been compared and the diagnostic accuracy of HE4, CA 125 and the combinatory ROMA algorithm for epithelial ovarian cancer (EOC) has been evaluated. MethodsHE4 and CA125 were measured on sera obtained from 259 women (73 healthy, 90 with benign ovarian or adnexal diseases, 96 with EOC). The ARCHITECT CMIA HE4 assay was compared with the Fujirebio EIA HE4, and the risk for EOC by the combinatory ROMA algorithm (HE4+CA 125) was assessed with both HE4 assays. ResultsThe CMIA HE4 assay showed a good linearity (r>0.9998) and precision (interassay and total CVs <4%). The correlation with EIA HE4 was linear (r=0.994), with an average bias of 0.4%. By ROC curve analysis, the sensitivity for EOC at a fixed specificity of 90%, 95% and 99% was 89.6%, 84.4% and 79.2% by CMIA HE4, 84.4%, 83.3% and 79.2% by EIA HE4, 86.5%, 76.0% and 59.4% by CMIA CA125. The accuracy of the ROMA algorithm determined by CMIA or EIA HE4 was very similar (AUC 87.1% vs. 87.6%; p=n.s.) and greater in menopause. ConclusionsThe two HE4 assays showed a good correlation and similar clinical value, with a greater precision for CMIA. HE4 was more specific and accurate than CA125, supporting its use in addition to clinical and imaging criteria for the discrimination of benign from malignant ovarian lesions. The ROMA algorithm showed a good accuracy for discriminating women at high risk for EOC.