Decoding biological age from face photographs using deep learning.

Because humans age at different rates, a person's physical appearance may yield insights into their biological age and physiological health more reliably than their chronological age. In medicine, however, appearance is incorporated into medical judgments in a subjective and non-standardized fashion. In this study, we developed and validated FaceAge, a deep learning system to estimate biological age from easily obtainable and low-cost face photographs. FaceAge was trained on data from 58,851 healthy individuals, and clinical utility was evaluated on data from 6,196 patients with cancer diagnoses from two institutions in the United States and The Netherlands. To assess the prognostic relevance of FaceAge estimation, we performed Kaplan Meier survival analysis. To test a relevant clinical application of FaceAge, we assessed the performance of FaceAge in end-of-life patients with metastatic cancer who received palliative treatment by incorporating FaceAge into clinical prediction models. We found that, on average, cancer patients look older than their chronological age, and looking older is correlated with worse overall survival. FaceAge demonstrated significant independent prognostic performance in a range of cancer types and stages. We found that FaceAge can improve physicians' survival predictions in incurable patients receiving palliative treatments, highlighting the clinical utility of the algorithm to support end-of-life decision-making. FaceAge was also significantly associated with molecular mechanisms of senescence through gene analysis, while age was not. These findings may extend to diseases beyond cancer, motivating using deep learning algorithms to translate a patient's visual appearance into objective, quantitative, and clinically useful measures.

Machine learning for normal tissue complication probability prediction: Predictive power with versatility and easy implementation.

Background and purpose: A popular Normal tissue Complication (NTCP) model deployed to predict radiotherapy (RT) toxicity is the Lyman-Burman Kutcher (LKB) model of tissue complication. Despite the LKB model's popularity, it can suffer from numerical instability and considers only the generalized mean dose (GMD) to an organ. Machine learning (ML) algorithms can potentially offer superior predictive power of the LKB model, and with fewer drawbacks. Here we examine the numerical characteristics and predictive power of the LKB model and compare these with those of ML. Materials and methods: Both an LKB model and ML models were used to predict G2 Xerostomia on patients following RT for head and neck cancer, using the dose volume histogram of parotid glands as the input feature. Model speed, convergence characteristics and predictive power was evaluated on an independent training set. Results: We found that only global optimization algorithms could guarantee a convergent and predictive LKB model. At the same time our results showed that ML models remained unconditionally convergent and predictive, while staying robust to gradient descent optimization. ML models outperform LKB in Brier score and accuracy but compare to LKB in ROC-AUC. Conclusion: We have demonstrated that ML models can quantify NTCP better than or as well as LKB models, even for a toxicity that the LKB model is particularly well suited to predict. ML models can offer this performance while offering fundamental advantages in model convergence, speed, and flexibility, and so could offer an alternative to the LKB model that could potentially be used in clinical RT planning decisions.

International consensus on radiotherapy in metastatic non-small cell lung cancer.

Background: Lung cancer is the leading cause of cancer-related death worldwide, with non-small cell lung cancer (NSCLC) accounting for most cases. While radiotherapy has historically served as a palliative modality in metastatic NSCLC, considerable advances in its technology and the continuous development of cutting-edge therapeutic agents, such as targeted therapy and immune checkpoint inhibitors (ICIs), are increasing its role in the multi-disciplinary management of the disease. Methods: International radiotherapy experts were convened to consider and reach consensuses on the clinical utilities of radiotherapy in metastatic NSCLC, with the aim to provide patient-focused, up to date, evidence-based, recommendations to assist cancer specialists in the management of patients with metastatic NSCLC worldwide. Results: Timely radiotherapy can offer rapid symptom alleviation and allow subsequent aggressive treatment approaches in patients with heavy tumor burden and/or oncologic emergencies. In addition, appropriate incorporation of radiotherapy as concurrent, consolidation, or salvage therapy makes it possible to achieve long-term survival, or even cure, for patients with oligo-metastatic disease. Cranial radiotherapy plays an important role in the management of brain metastasis, potentially augmenting the response and prolonging survival associated with targeted agents and ICIs. However, key questions remain, such as the appropriate choice of radiation techniques, optimal sequence of systemic therapies and radiotherapy, and optimal patient selection for such combination strategies. Although a strong rationale for combining radiotherapy and ICIs exists, its optimal parameters in this setting remain to be established. Conclusions: In the modern era, radiotherapy serves not only as a palliative tool in metastatic NSCLC, but also plays active roles in patients with oligo-focal disease, CNS metastasis and receiving ICIs.

Repeat reirradiation of the spinal cord: multi-national expert treatment recommendations.

BACKGROUND: Improved survival of patients with spinal bone metastases has resulted in an increased number of referrals for retreatment and repeat reirradiation. METHODS: A consortium of expert radiation oncologists (RO) has been established with the aim of providing treatment recommendations for challenging clinical scenarios for which there are no established guidelines. In this case, a patient developed local progression of a T5 vertebral lesion after two prior courses of palliative radiotherapy (time interval >12 months, assumed cumulative biologically equivalent dose in 2­Gy fractions [EQD2] for spinal cord [alpha/beta 2 Gy] 75 Gy). Expert recommendations were tabulated with the aim of providing guidance. RESULTS: Five of seven RO would offer a third course of radiotherapy, preferably with advanced techniques such as stereotactic radiotherapy. However, the dose-fractionation concepts were heterogeneous (3-20 fractions) and sometimes adjusted to different options for systemic treatment. All five RO would compromise target volume coverage to reduce the dose to the spinal cord. Definition of the spinal cord planning-organ-at-risk volume was heterogeneous. All five RO limited the EQD2 for spinal cord. Two were willing to accept more than 12.5 Gy and the highest EQD2 was 19 Gy. CONCLUSIONS: The increasing body of literature about bone metastases and spinal cord reirradiation has encouraged some expert RO to offer palliative reirradiation with cumulative cord doses above 75 Gy EQD2; however, no consensus was achieved. Strategies for harmonization of clinical practice and development of evidence-based dose constraints are discussed.

Postoperative radiotherapy for lung cancer: Is it worth the controversy?

INTRODUCTION: The role of postoperative radiation therapy (PORT) in patients with completely resected non-small cell lung cancer (NSCLC) with pathologically involved mediastinal lymph nodes (N2) remains unclear. Despite a reduction of local recurrence (LR), its effect on overall survival (OS) remains unproven. Therefore we conducted a review of the current literature. METHODS: To investigate the benefit and safety of modern PORT, we identified published phase III trials for PORT. We investigated modern PORT in low-risk (ypN0/1 and R0) and high-risk (ypN2 and/or R1/2) patients with stage III-N2 NSCLC treated with induction chemotherapy and resection. RESULTS: Seventeen phase III trials using PORT were selected. Of all PORT N2 studies, 4 were eligible for evaluation of LR, all in high-risk patients only. In these high-risk patients receiving PORT, the mean LR rate at 5years was 20.9% (95% CI 16-24). Two trials were suitable to assess LR rates after chemotherapy and surgery without PORT. In these low-risk patients, the mean 5-year LR was 33.1% (95% CI 27-39). No significant difference in non-cancer deaths between PORT vs. non-PORT patients was observed in N2 NSCLC. CONCLUSION: PORT is worth the controversy because data illustrate that PORT may increase the OS. However, prospective randomized trials are needed to verify this.

The cost-effectiveness of particle therapy in non-small cell lung cancer: exploring decision uncertainty and areas for future research.

PURPOSE: To review and synthesize all available evidence in order to explore the cost-effectiveness of particle therapy (carbon-ions, protons) compared to the best current treatments for non-small-cell lung cancer (NSCLC), and the value of additional research. The present study focuses on stage I NSCLC, as no data is available for more advanced stages. METHODS: A probabilistic decision-analytic Markov model was constructed to synthesize all available evidence. Comparative treatments were carbon-ions, protons, conventional radiotherapy (CRT) and stereotactic radiotherapy (SBRT) for inoperable stage I NSCLC; and carbon-ions and SBRT for operable stage I NSCLC. The expected value of perfect information (EVPI) was calculated to support research decisions. RESULTS: For inoperable stage I NSCLC, carbon-ion therapy costed euro 67.257 per quality-adjusted-life-year gained compared to SBRT. Both treatments dominated protons and CRT. Considerable uncertainty surrounded these results, resulting in a high EVPI. For operable stage I NSCLC SBRT dominated carbon-ion therapy. CONCLUSIONS: Due to the considerable uncertainty in stage I NSCLC, and the lack of data on more advanced stages, it is recommended not to adopt particle therapy as standard treatment in NSCLC yet. More evidence is needed to reduce the decision uncertainty and to support evidence-based treatment decisions. It might be worthwhile to invest in a particle facility for clinical research. Future research should also weigh the investment risk, value of information and costs of delay.

Transition from a simple to a more advanced dose calculation algorithm for radiotherapy of non-small cell lung cancer (NSCLC): implications for clinical implementation in an individualized dose-escalation protocol.

BACKGROUND AND PURPOSE: To investigate the clinical consequences of the transition from a simple convolution algorithm (CA) to a more advanced superposition dose calculation algorithm (SA) in an individualized isotoxic dose-escalation protocol for NSCLC patients. MATERIAL AND METHODS: First, treatment plans designed according to ICRU50-criteria using the CA were recalculated using the SA, for 16 patients. Next, two additional plans were designed for each patient using only the SA: one with 95%-isodose coverage (ICRU50-criteria), the other allowing PTV coverage with 90%-isodose at the lung side. PTV dose was escalated to a maximum dose of 79.2Gy or lower when limited by either a mean lung dose (MLD) of 19Gy or a maximum spinal cord dose of 54Gy. Equivalent uniform doses (EUD) in the PTV were compared. RESULTS: Recalculation of the CA plans using the SA, showed PTV underdosage in the CA plans: the median PTV EUD was 61.3Gy (range 44.9-80.4Gy) and 55.5Gy (43.9-76.8Gy), for CA and SA, respectively (p<0.001). Redesigning plans using the SA resulted in an almost identical PTV EUD of 55.1Gy (43.7-79.2Gy). For the subgroup (N=9) with MLD as dose-limiting factor a gain in PTV EUD of 2.7+/-1.8Gy (p=0.008) was achieved using the 90%-isodose coverage plan. CONCLUSIONS: Plans calculated using the CA caused large PTV underdosage. Plans designed using the SA often lead to lower maximum achievable tumour doses due to higher MLD values. Allowing somewhat relaxed PTV coverage criteria increased the PTV dose again for MLD restricted cases. Consequently, in clinics where isotoxic individual dose-escalation is applied, implementation of an SA should be accompanied by accepting limited PTV underdosage in patients with MLD as the dose-limiting factor.

Particle therapy in lung cancer: where do we stand?

BACKGROUND: From a theoretical point of view, charged particles should lead to superior results compared to photons. In this review, we searched for clinical evidence that protons or C-ions are really beneficial to patients with lung cancer. METHODS: A systematic literature review based on an earlier published comprehensive review was performed and updated until November 1st 2007. RESULTS: Ten fully published series, all dealing with non-small cell lung cancer (NSCLC), mainly stage I, were identified. No phase III trials were found. On proton therapy, 2-5 year local tumor control rates varied between 87% and 57%. The 2 year/5 year overall survival and 2 year/5 year cause specific survival varied between 31-74%/23% and 58-86%/46%, respectively. Late side effects were observed in about 10% of the patients. For C-ion therapy, the local tumor control rate was 77%, while 95% when using a hypofractionated radiation schedule. The 5 year overall survival and cause specific survival rates were 42% and 60%, respectively. Slightly better results were reported when using hypofractionation, 50% and 76%, respectively. The reported late side effects for C-ions were 4%. CONCLUSION: The results with charged particles, at least for stage I disease, seem to be promising. A gain can be expected in reduction of late side effects, especially after treatment with C-ions. Available data demonstrate that particle therapy in general is a safe and feasible treatment modality. Although current results are promising, more evidence is required before particle therapy can become internationally the standard treatment for (subsets of) lung cancer patients.